Merged Public

Designing for Humans in Autonomous Systems: 在自主系统中为人类设计:
Military Applications 军事应用

由Michael J.Barnes，Jessie Y.C.Chen，Florian Jentsch，Tal Oron-Gila
d，Elizabeth Redden，Linda Elliott和Arthur W.Evans III
by Michael J. Barnes, Jessie Y. C. Chen, Florian Jentsch, Tal Oron-Gilad,
Elizabeth Redden, Linda Elliott, and Arthur W. Evans III
ARL-TR-6782 January 2014 ARL-TR-6782 January 2014
Approved for public release; distribution is unlimited. 批准公开发布;分发是无限的.

NOTICES NOTICES
Disclaimers Disclaimers
除非其他授权文件指定，否则本报告中的调查结果不得解释为军队职位的官方部门。
The findings in this report are not to be construed as an official Department of the Army position unless
so designated by other authorized documents.
引用制造商或商号并不构成对其使用的正式认可或批准。
Citation of manufacturer’s or trade names does not constitute an official endorsement or approval of the
use thereof.
当不再需要此报告时，请销毁此报告。不要把它还给发端人。
Destroy this report when it is no longer needed. Do not return it to the originator.
Army Research Laboratory 陆军研究实验室
Aberdeen Proving Ground, MD 21005-5425 香港仔试验场,MD21005-5425
ARL-TR-6782 January 2014 ARL-TR-6782 January 2014
Designing for Humans in Autonomous Systems: 在自主系统中为人类设计:

Military Applications 军事应用
Michael J. Barnes, Jessie Y. C. Chen, Elizabeth Redden, Linda Elliott, and Michael J.Barnes，Jessie Y.C.Chen，Elizabeth Redden，Linda Elliott和
Arthur W. Evans III 亚瑟*W*埃文斯三世
Human Research and Engineering Directorate, ARL 人类研究及工程局
Florian Jentsch Florian Jentsch

University of Central Florida 中佛罗里达大学
Tal Oron-Gilad Tal Oron-Gilad

Ben-Gurion University of the Negev 内盖夫本-古里安大学
批准公开发布;分发是无限的.
Approved for public release; distribution is unlimited.
Form Approved 批准表格
REPORT DOCUMENTATION PAGE
OMB No. 0704-0188
报告文档页面 OMB No. 0704-0188
Public reporting burden for this collection of information is estimated to average 1 hour per response, including the time for reviewing instructions, searching existing data sources, gathering 这种收集信息的公共报告负担估计平均每个答复1小时包括审查指示、搜索现有数据源、收集和维护所需数据以及完成和审查收集信息的时间。向国防部、华盛顿总部服务部、信息业务和报告局
and maintaining the data needed, and completing and reviewing the collection information. Send comments regarding this burden estimate or any other aspect of this collection of information, （0704-0188）、杰斐逊*戴维斯公路1215号，阿灵顿，VA22202-4302号，发送关于这一负担估计或这一信息收集的任何其他方面的意见，包括减轻负担的建议。受访者应注意，尽管有任何其
including suggestions for reducing the burden, to Department of Defense, Washington Headquarters Services, Directorate for Information Operations and Reports (0704-0188), 1215 Jefferson 他法律规定，如果信息收集没有显示当前有效的OMB控制号码，任何人都不得因未能遵守信息收集而受到任何处罚。请勿将表格交回上述地址。
Davis Highway, Suite 1204, Arlington, VA 22202-4302. Respondents should be aware that notwithstanding any other provision of law, no person shall be subject to any penalty for failing to
comply with a collection of information if it does not display a currently valid OMB control number.
PLEASE DO NOT RETURN YOUR FORM TO THE ABOVE ADDRESS.
1. REPORT DATE (DD-MM-YYYY) 2. REPORT TYPE 3. DATES COVERED (From - To) 1. 报告日期(DD-MM-YYYY)2014年 2. 报告类型最终 3. 涵盖日期（自）2009年10月至2012年
1月 12月
January 2014 Final October 2009–December 2012
4. TITLE AND SUBTITLE 5a. CONTRACT NUMBER
4. 标题和副标题自主系统中人类设计：军事应用 5A.合约编号
Designing for Humans in Autonomous Systems: Military Applications 5B.补助金编号
5b. GRANT NUMBER
5c. PROGRAM ELEMENT NUMBER 5C.程序元素编号
6. AUTHOR(S) 5d. PROJECT NUMBER

6. 作者Michael J.Barnes,Jessie Y.C.Chen,Florian Jentsch,*Tal Oron-Gilad,†Eli 5d.项目编号
5E.任务编号
Michael J. Barnes, Jessie Y. C. Chen, Florian Jentsch,* Tal Oron-Gilad,† Elizabeth zabeth
5e. TASK NUMBER
Redden, Linda Elliott, and Arthur W. Evans III Redden，Linda Elliott和Arthur W.Evans III
5f. WORK UNIT NUMBER 5F.工作单位编号
7. PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES) 8. PERFORMING ORGANIZATION 7. 执行机构名称和地址美国陆军研究实验室 8. 执行组织报告编号

REPORT NUMBER
U.S. Army Research Laboratory
ATTN: RDRL-HRM-AY ARL-TR-6782 ATTN: RDRL-HRM-AY ARL-TR-6782
Aberdeen Proving Ground, MD 21005-5425 香港仔试验场,MD21005-5425
9. SPONSORING/MONITORING AGENCY NAME(S) AND ADDRESS(ES) 10. SPONSOR/MONITOR’S ACRONYM(S) 9. 赞助/监察机构名称及地址 10. SPONSOR/MONITOR’S ACRONYM(S)
11. SPONSOR/MONITOR'S REPORT 11. SPONSOR/MONITOR'S REPORT

NUMBER(S) NUMBER(S)
12. DISTRIBUTION/AVAILABILITY STATEMENT

12. 发布/可用性声明批准公开发布；发布是无限
Approved for public release; distribution is unlimited. 的。
13. SUPPLEMENTARY NOTES
13. 补充说明*中佛罗里达大学心理学系，3100技术Pkwy。，斯特。 100,奥兰多,FL32826†内盖夫本古里安大学
*
Department of Psychology, University of Central Florida, 3100 Technology Pkwy., Ste. 100, Orlando, FL 32826 工业工程与管理系,P.O.Box653,Beer-Sheva,以色列,84105
†
Department of Industrial Engineering and Management, Ben-Gurion University of the Negev, P.O. Box 653, Beer-Sheva,
Israel, 84105
14. ABSTRACT
14. 摘要本报告的目的是回顾美国陆军研究实验室（ARL）赞助的关于人类在未来自治系统中的作用的研究，并得出旨
The purpose of this report is to review U.S. Army Research Laboratory (ARL)-sponsored research on the human’s role in future 在促进人类/自治协作的设计指南。该研究是作为一个更大的军队计划的一部分进行的，该计划是在复杂环境中进行侦
autonomous systems and to derive design guidelines to foster human/autonomy collaboration. The research was conducted as 察的无人系统的安全操作（来源），重点是为城市应用开发安全自治。人类自主设计研究包括代理可靠性、控制范围
part of a larger Army program, Safe Operations for Unmanned systems for Reconnaissance in Complex Environments (SOURCE), 、安全问题、个体差异、培训、功能分配以及评估高级接口解决方案的现场实验结果。本报告的主要部分包括（1）
that focused on developing safe autonomy for urban applications. The human-autonomy design research encompasses agent 自主和智能代理，（2）RoboLeader，（3）自主系统的安全，（4）自然界面和（5）使用无人驾驶车辆图像的情况理
reliability, span of control, safety issues, individual differences, training, function allocation, and results from field experiments 解。在每个部分之后，总结了结果的含义，以制定将人类纳入自主军事系统的设计指南。
evaluating advanced interface solutions. The main sections of this report cover (1) autonomy and intelligent agents, (2) RoboLeader,
(3) safety for autonomous systems, (4) naturalistic interfaces, and (5) situation understanding using unmanned vehicle imagery.
After each section, implications of the results are summarized to develop design guidelines for incorporating humans into
autonomous military systems.
15. SUBJECT TERMS
15. 学科术语人机交互,自主,小型机器人,自主的人为因素,自主无人地面车辆
human robot interaction, autonomy, small robots, human factors of autonomy, autonomous unmanned ground vehicles
17. LIMITATION 18. NUMBER 19a. NAME OF RESPONSIBLE PERSON 17. LIMITATION 18. NUMBER 19a.负责人姓名Michael J.Barnes
16. SECURITY CLASSIFICATION OF: OF ABSTRACT OF PAGES 16. 安全分类: 摘要页数
Michael J. Barnes
a. REPORT b. ABSTRACT c. THIS PAGE 19b. TELEPHONE NUMBER (Include area code) A.报告未分 B.抽象的未 C.本页未分类 19B.电话号码(包括区号)520-538-4702
Unclassified Unclassified Unclassified UU 36 520-538-4702 类分类 UU 36
Standard Form 298 (Rev. 8/98) 标准表格298(Rev.8/98)
Prescribed by ANSI Std. Z39.18 ANSI Std规定。 Z39.18
ii ii
Contents Contents
List of Figures iv 数字一览表 iv
1. Introduction 1 1. Introduction 1
2. Autonomy and Intelligent Agents 2 2. 自治和智能代理 2
3. RoboLeader and Control of Multiple Systems 3 3. 多个系统的RoboLeader和控制 3
4. Safety and Levels of Autonomy 6 4. 安全和自主水平 6
5. Ft. Benning and Ft. Leonard Wood Field Experiments: Progressive Autonomy and 5. 英尺。本宁和英国《金融时报》。伦纳德*伍德野外实验：渐进自治和
Intuitive Interfaces 9 直观的界面 9
6. Situation Understanding and Decision Support 14 6. 情况理解和决策支持 14
7. Conclusions and Future Research 19 7. 结论和未来研究 19
8. References 22 8. References 22
List of Symbols, Abbreviations, and Acronyms 27 符号、缩写和首字母缩写词列表 27
Distribution List 29 分发名单 29
iii iii
List of Figures 数字一览表
Figure 1. Simulation scene showing text window, robot instrument gauges, the robot location 图1。仿真场景显示文本窗口、机器人仪器仪表、机器人位置图、机器人相机视图的四个
map, four windows of robot camera views, and a larger display of scene as viewed by 窗口以及机器人4所看到的更大的场景显示。 .............................................................................................................................
Robot 4. ......................................................................................................................................4 .........4
Figure 2. Views of Middle Eastern urban setting as viewed by operators in the control room. ......6 图2。控制室内操作员所看到的中东城市环境景观。 ......6
Figure 3. Interaction between reliability and LOA. .........................................................................8 图3。可靠性和LOA之间的相互作用。 .........................................................................8
Figure 4. SPAWAR robot. .............................................................................................................10 图4。 SPAWAR机器人. .............................................................................................................10
Figure 5. The haptic manipulator used with stereovision for the Talon robot experiment............13 图5。触觉机械手与立体视觉一起用于爪机器人实验. ...........13
Figure 6. Ben-Gurion University combined view display. ............................................................15 图6。本-古里安大学联合视图显示。 ............................................................15
Figure 7. An example of the Warfighter machine interface (WMI) showing both the short- 图7。战斗机机器接口（WMI）的一个例子显示了这两个短
term (green) and long-term (blue) operator aids. .....................................................................17 长期（绿色）和长期（蓝色）操作员辅助工具。 .....................................................................17
iv iv
1. Introduction 1. Introduction
The possible applications of autonomy in the military are becoming ubiquitous for unmanned 自治在军事上的可能应用正在变得无处不在，用于无人机，地面，海军和水下设置（Che
aerial, ground, naval, and underwater settings (Chen et al., 2011a). The advantages for military n et al.，2011a)。军事系统的优势包括减少人力，增加生存能力，性能改进以及载人系
systems include reduced manpower, increased survivability, performance improvements, and 统无法执行的战术（Barnes和Evans，2010）。在未来，自主系统可能能够以速度行驶，
tactics that would be impossible for manned systems to perform (Barnes and Evans, 2010). In the 并在人类控制系统不可能的情况下运行。目前，大多数无人系统仍然需要人工操作，因
future, autonomous systems may be capable of travelling at speeds and operating under 此对人员减少没有显着的影响。例如，陆军的战术无人机系统（UAS）基本单位需要两个
地面站，22人员，包括一名军官和一名准尉。此外，维修由民用承包商提供（Zapotoczn
circumstances impossible for human-controlled systems. Currently, most unmanned systems still
y，2007）。同样，空军在飞行员操作长耐力UAS全球鹰等系统时也存在重大人员配备问
need human operators and thus do not have a significant effect on personnel reduction. For
题，机组规模也是人类控制无人地面车辆（Ugv）的关键问题（Barnes和Evans，2010;Mi
example, the Army’s tactical unmanned aerial system (UAS) basic unit requires two ground tchell和Chen，2006;Weiss，2011）。改进的自主性对于减少人员，同时保持可接受的
stations with 22 personnel, including an officer and a warrant officer. In addition, maintenance is 性能和效率是必要的（Chen et al.，2011a)。由于无人系统的能力正在迅速提高，人类控
supplied by civilian contractors (Zapotoczny, 2007). Similarly, the Air Force also has significant 制系统和至少部分自治系统之间的性能差距将在不太遥远的未来向自治倾斜（Chen和Bar
manning issues when Airmen operate systems such as the long endurance UAS Global Hawk, nes，in press;Weiss，2011）。这份报告的目的是调查随着自主成为军事现实而与人类
and crew size is also a crucial issue for human control of unmanned ground vehicles (UGVs) 相关的设计含义(Barnes et al., 2011).
(Barnes and Evans, 2010; Mitchell and Chen, 2006; Weiss, 2011). Improved autonomy is
necessary to reduce personnel while maintaining acceptable performance and efficiency (Chen et
al., 2011a). Because the capabilities of unmanned systems are improving rapidly, the
performance gap between human-controlled systems and at least partially autonomous systems
will tilt toward autonomy in the not too distant future (Chen and Barnes, in press; Weiss, 2011).
The objective of this report is to investigate the human-related design implications as autonomy
becomes a military reality (Barnes et al., 2011).
We focus on results from the human-robot interaction (HRI) program that supported a larger 我们专注于人机交互（HRI）程序的结果，该程序支持更大的
Army technology objective, Safe Operations for Unmanned systems for Reconnaissance in 陆军技术目标，在复杂环境中进行侦察的无人系统的安全操作（来源）。 SOURCE研究
Complex Environments (SOURCE). SOURCE researchers successfully demonstrated improved 人员成功地证明了在城市地形中导航的大型Ugv的改进自主性，同时避开行人和其他障碍
autonomy for large UGVs navigating in urban terrain while avoiding pedestrians and other 物。此外，这项研究还证明了小型UGVs（SUGVs）自主绘制内部空间的有效性，以及协
obstacles. In addition, this research also demonstrated the effectiveness of small UGVs (SUGVs) 调小型UGVs和UASs以测量建筑物的叛乱指标的潜力。与人相关的研究集中在自主性的设
mapping interior spaces autonomously and the potential of coordinating small UGVs and UASs 计含义和各种高级界面设计的效用。
to surveil buildings for insurgency indicators. The human-related research focused on the design
implications of autonomy and the utility of various advanced interface designs.
In this report, we examine various theoretical papers supporting the program as well as empirical 在这份报告中，我们审查了支持该计划的各种理论论文，以及来自美国陆军研究实验室
results from in-house researchers from the U.S. Army Research Laboratory’s (ARL’s) Human （ARL）人类研究与工程局（HRED）和合同研究的内部研究人员的实证结果。人类自主
Research and Engineering Directorate (HRED) and contracted research. The human-autonomy 设计研究包括代理可靠性、控制范围、个体差异、培训、功能分配以及评估高级接口解决
design research encompasses agent reliability, span of control, individual differences, training, 方案的现场实验结果。这份报告的五个主要部分包括自治和智能代理、RoboLeader、自
function allocation, and results from field experiments evaluating advanced interface solutions. 治系统的安全、自然界面和使用紫外线图像的情况理解。
The five main sections of this report cover autonomy and intelligent agents, RoboLeader, safety
for autonomous systems, naturalistic interfaces, and situation understanding using UV imagery.
1 1
After each section, implications of the results are summarized in order to develop design 在每个部分之后，总结了结果的含义，以便制定将人类纳入自主军事系统的设计
guidelines for incorporating humans into autonomous military systems. 指南。
2. Autonomy and Intelligent Agents 2. 自治和智能代理
A number of theoretical and review papers were written to guide the empirical research and to 撰写了一些理论和评论论文来指导实证研究并制定扩展内部研究边界的指导方针（Barne
develop guidelines that extend the boundaries of in-house research (Barnes and Chen, 2012; s和Chen，2012;Chen和Barnes，in press;Chen et al.，2011a;2011b)。在这些论文中，C
Chen and Barnes, in press; Chen et al., 2011a; 2011b). In these papers, Chen and colleagues hen及其同事讨论了适应未来军事环境所需的范式变化，其中数百辆载人和无人驾驶战车
discuss paradigm changes necessary to adjust to future military environments wherein hundreds 将协同运行。他们得出的结论是，随着自主性的提高，人类的角色将不可避免地转变为
of manned and unmanned combat vehicles will operate in concert. They concluded that as 对多个无人系统的监督，并且手动控制将成为仅在紧急情况下使用的回退模式（陈等人
。, 2007). 此外，他们的结论强调了开发新的显示概念以提高态势感知（SA）和增强信任
autonomy improved, the human’s role will inevitably change to supervision of multiple unmanned
的重要性，因为现代战斗的内在复杂性需要对多个自治系统进行监督。具体而言，随着
systems, and that manual control will become a fallback mode used only in cases of emergencies
要控制的资产数量超过人类的注意力范围，直接监督控制变得低效（Miller，1956）。
(cf. Chen et al., 2007). Additionally, their conclusions stress the importance of developing new
display concepts to improve situation awareness (SA) and to enhance trust because of the inherent
complexity of modern combat entailing supervision of multiple autonomous systems. Specifically,
direct supervisory control becomes inefficient as the number of assets to be controlled surpasses
the human’s attention span (Miller, 1956).
To address these issues, Chen developed concepts that are similar to executive supervision 为了解决这些问题，陈提出了类似于执行监督的概念。 Barnes和Grossman，1985;Chen
(cf. Barnes and Grossman, 1985; Chen and Barnes, 2012a; 2012b; in press). The analogy can 和Barnes，2012a;2012b;in press）。这个比喻可以最好地描述为店长和行政人员之间的
best be described as the difference between the shop foremen and the executive. The foreman 区别。工头直接监督工人（每个工人都有指定的自治程度），执行官直接与工头合作制
directly supervises the workers (who each have a specified degree of autonomy), and the 定战略，是重要变化的最终决策者。 Chen及其同事开发了使用软件代理允许单个操作员
executive works directly with the foreman to develop strategies and is the final decision maker 在多任务环境中控制多个资产的理论基本原理和经验证据（Chen和Barnes，2012a;2012
b;in press）。代理人被定义为"观察和对环境采取行动并指导其活动以实现目标的自治实
for important changes. Chen and colleagues developed the theoretical rationale and empirical
体"（Russell和Norvig，2009，p.34）。执行范式的优势在于，人（执行）可以自由地维
evidence for using software agents to allow a single operator to control multiple assets in a
护整体SA并做出战术决策，而代理（中间主管）则向操作员发出可能出现的问题的警报
multitasking environment (Chen and Barnes, 2012a; 2012b; in press). An agent is defined as “an ，并在半自治资产需要干预时提出缓解问题的行动方案。
autonomous entity which observes and acts upon an environment and directs its activity towards
achieving goals” (Russell and Norvig, 2009, p. 34). The advantage of an executive paradigm is
that the human (executive) is freed to maintain overall SA and to make tactical decisions while
the agent (intermediate supervisor) alerts the operator to possible problems and suggests courses
of action to mitigate problems when semi-autonomous assets require intervention.
Chen’s agents are an example of mixed-initiative systems wherein both the agent and the human 陈的代理人是混合主动系统的一个例子，其中代理人和人都可以煽动决策，但后者具有
can instigate decisions, but the latter has ultimate decision authority (Goodrich, 2010). They are 最终决策权（Goodrich，2010）。它们与自适应系统形成对比，其中控制在指定条件下
contrasted with adaptive systems where control passes between operators and intelligent software 在操作员和智能软件之间传递，以及可调节系统，其中人类决定要引出哪些自主行为（Ba
under specified conditions and adjustable systems wherein the human decides which of the rnes et al.，2006;Chen和Barnes，in press;Miller和Parasuraman，2007）。
repertoire of autonomous behaviors to elicit (Barnes et al., 2006; Chen and Barnes, in press; Miller
and Parasuraman, 2007).
2 2
In their most recent articles, Chen and Barnes (in press) concluded that a mixed-initiative 在他们最近的文章中，Chen和Barnes（in press）得出结论，由个人代理通知的混合主动
architecture informed by a personal agent is the most flexible control structure for executive 架构是执行监督最灵活的控制结构。这样的体系结构允许人类对多个自治系统有最终的
supervision. Such architectures allow the human to have ultimate control over multiple 控制，而代理作为人类和更专业的自治系统之间的接口。他们在审查相关文献和内部研
autonomous systems, while the agent acts as the interface between humans and more specialized 究的基础上制定了以下指导方针:
autonomous systems. They developed the following guidelines based on a review of both the
pertinent literature and in-house research:
• Agent/human interaction needs to be flexible. The user interface should support • 代理/人的交互需要灵活。用户界面应支持双向通信和控制结构，以实现快速变化
bidirectional communications and control structures to effect rapid change. The system 。系统应该能够适应操作员的工作负载，并允许代理在操作员指定的条件下自主
should be able to adjust to operator workload and allow agents to act autonomously under 行动。
operator-specified conditions.
• The user interface must enable the operator’s ultimate decision authority. The mechanism • 用户界面必须启用操作员的最终决策权限。确保人的权威的机制需要嵌入代理体
for ensuring human authority needs to be embedded in the agent architecture (e.g., mixed- 系结构（例如，混合主动系统）中。
initiative systems).
• Automation transparency is essential. Lee and See’s (2004) 3P’s (purpose, process, and • 自动化透明度至关重要。 Lee And See's(2004)3P's(purpose,process,and performa
performance) as well as the history of the system’s 3P’s should be presented to the operator nce)以及系统的3p's的历史应该以简化的形式呈现给操作员，例如集成的图形显示。
in a simplified form, such as integrated graphical displays. The user interface must support 用户界面必须支持操作员对代理的行为和任务环境的理解以及中断后的有效任务恢
operator understanding of the agent’s behavior and the mission environment as well as 复。
effective task resumption after interruptions.
• Visualization and training techniques should act as enablers of human-agent collaboration. • 可视化和培训技术应该成为人机协作的推动者。
Appropriate human-agent trust (Lee and See, 2004) can be reinforced by both training and 适当的人-代理信任（Lee和See，2004）可以通过培训和可视化方法来加强。特
visualization methods. Specific visualization techniques (e.g., augmented reality) have 定的可视化技术（例如，增强现实）已被证明在代理技术最有益的复杂环境类型中
proven to be particularly useful in improving SA in the type of complex environments 在改进SA方面特别有用。操作员应该接受培训，以了解系统的3P。
where agent technology is most beneficial. Operators should be trained to understand the
system’s 3P’s.
• Human individual differences must be part of the human/agent design process. This • 人的个体差异必须是人/代理设计过程的一部分。该准则可以通过界面设计、选
guideline can be accomplished by interface design, selection, training, and/or designing 择、培训和/或设计对人类个体差异敏感的代理来实现。
agents that are sensitive to individual differences among humans.
3. RoboLeader and Control of Multiple Systems 3. 多个系统的RoboLeader和控制
Chen and colleagues simulated vehicular combat situations wherein the operator was burdened Chen及其同事模拟了车辆战斗情况，其中操作员背负着多任务要求以及监督多个自治系
with multitasking requirements as well as supervising multiple autonomous systems (Chen, 统（Chen，2011;Chen和Barnes，2012a;2012b）。完全自治的系统在这种环境中是不
2011; Chen and Barnes, 2012a; 2012b). Completely autonomous systems would not be practical 实用的，因为自治会限制战术灵活性并带来安全风险，而多个自治系统的监督控制则带来
in this environment because autonomy would limit tactical flexibility and pose safety risks, while 了自己的问题，如自满效应和短期记忆限制。陈等人。 (2011b)介绍
supervisory control of multiple autonomous systems introduced its own problems, such as
complacency effects and short-term memory limitations. Chen et al. (2011b) introduced the
3 3
concept of employing RoboLeader, an intelligent agent, to assess the current state of multiple 使用智能代理RoboLeader的概念来评估多个系统的当前状态，提出算法解决方案，并仅
systems, suggest algorithmic solutions, and execute them only when given permission by the 在操作员许可的情况下执行它们。其优点是操作员可以维护SA并处理其他任务，而Rob
operator. The advantage is that the operator could maintain SA and attend to other tasks while oLeader将作为下属船员，专注于机器人资产的当前状态。
RoboLeader would act as a subordinate crew member focused on the current state of the robotic
assets.
The first experiment was a proof-of-concept for RoboLeader (Chen and Barnes, 2012a). Humans 第一个实验是RoboLeader的概念验证（Chen和Barnes，2012a）。与RoboLeader合作
working with RoboLeader were able to successfully reroute up to eight robots more rapidly than 的人类能够成功地在遇到意外障碍时比手动条件更快地重新路由多达八个机器人。在第
manual conditions when unexpected obstacles were encountered. In the second experiment, 二个实验中，RoboLeader的可靠性（60％和90％）和错误类型（容易发生误报[FAP]和容
reliability of RoboLeader (60% and 90%) and type of errors (false alarm prone [FAP] and miss 易发生误报[MP]）进行了参数化。图1显示了Mix Initiative实验（MIX）仿真环境，其中
prone [MP]) were varied parametrically. Figure 1 shows the Mix Initiative Experimental (MIX) 包括（1）用于机器人重新路由的地图显示，（2）显示机器人视图的小窗口，（3）用于
目标识别的较大窗口，（4）仪表板，以及（5）以前的研究（Wickens等。（2010年）
simulation environment that includes (1) a map display for robot rerouting, (2) small windows
指出，与MP加权系统相比，高FAP警报对整体性能的危害更大，因为"哭狼效应"导致运营
showing views from the robots, (3) a larger window for target identification, (4) instrument
商对FAP警报失去信心。
panels, and (5) a text window for RoboLeader to communicate with operators. Previous research
(Wickens et al., 2010) indicated that high FAP alerts were more deleterious to overall
performance than were MP-weighted systems because the “cry wolf effect” caused operators to
lose faith in FAP alerts.
Figure 1. Simulation scene showing text window, robot instrument gauges, the 图1。仿真场景显示文本窗口、机器人仪器仪表、机器人位置图、机器人
robot location map, four windows of robot camera views, and a larger 相机视图的四个窗口以及机器人4所看到的更大的场景显示。
display of scene as viewed by Robot 4.
Conversely, in the second study (Chen and Barnes, 2012a) for agents with 80% average error 相反，在第二项研究（Chen和Barnes，2012a）中，对于平均错误率为80％的代理，与
rates, the FAP conditions resulted in better overall performance for both target acquisition and MP代理相比，FAP条件导致目标获取和路由效率的整体性能更好。在这项研究中，由于
routing efficiency compared to the MP agents. In this study, the location of the robots could be 嵌入式地图显示的布局，可以轻松检查机器人的位置以获取FAP警报，而不是以前的研
checked easily for FAP alerts as opposed to previous research because of the layout of the 究。因此，这使得fap警报的合规性变得高效，因为目标获取的注意力切换相对容易。
embedded map display. As a result, this made compliance to FAP alerts efficient because of the 相比之下，MP代理在更大程度上干扰了操作员的表现，因为mp中的参与者
relative ease of attention switching for target acquisition. By way of contrast, the MP agent
interfered with the operator’s performance to a greater degree because participants in the MP
4 4
conditions had to continually check data on the map, thus drawing their attention away from the 条件必须不断检查地图上的数据，从而将他们的注意力从目标显示器上移开。这一结合
target displays. This conjuncture was supported by the SA measures indicating better 得到了SA措施的支持，这些措施表明MP条件下地图相关数据的性能更好，这表明操作员
performance on map-related data for MP conditions, suggesting that an operator’s focus on the 对地图显示的关注对其扫描性能不利。此外，由于个体差异也有显着的影响;例如，对注
map display is detrimental to their scanning performance. Also, there were significant effects due 意力控制能力高度自信的参与者具有更好的整体MP表现，表明他们转移注意力资源的能
to individual differences; for example, participants who were highly confident in their attentional 力。同样，更高水平的空间能力和游戏体验对目标性能产生了积极影响。
control abilities had better overall MP performance, indicating their ability to shift their
attentional resources. In the same vein, higher levels of spatial ability and gaming experience had
positive effects on targeting performance.
In another experiment (Chen and Barnes, 2012b), RoboLeader used more sophisticated 在另一个实验中（Chen和Barnes，2012b），RoboLeader使用更复杂的算法来引导四个
algorithms to direct four robots to entrap a moving vehicular target. Levels of autonomy (LOAs) 机器人夹带移动的车辆目标。自治水平（LOAs）随着可视化辅助的增加而变化。该实验
were varied with the addition of a visualization aid. The purpose of this experiment was to assess 的目的是评估RoboLeader代理在更动态的战斗环境中的有效性，其中目标和追求的机器
the effectiveness of the RoboLeader agent in a more dynamic combat environment in which both 人都在移动。有四个LOA条件：（1）手动，（2）混合，（3）与可视化混合，（4）与
the targets and the pursuing robots were moving. There were four LOA conditions: (1) manual, 可视化完全自动化。在混合条件下，人类操作员为追逐机器人选择了终点，RoboLeader
计算了一个最优解来捕获运动目标。可视化辅助显示了机器人的进度与捕获移动目标的
(2) hybrid, (3) hybrid with visualization, and (4) fully automated with visualization. For the
最佳解决方案的差异。总体而言，86％的参与者使用全自动化解决方案成功地实现了目
hybrid condition, the human operator chose end points for the pursuing robots, and RoboLeader
标，而96％的参与者使用混合解决方案（没有可视化）这样做，这表明在中佛罗里达大学
computed an optimal solution to entrap the moving target. The visualization aid displayed how （UCF）研究中可视化辅助对性能的影响很小；即使部分自治，地图显示上的原始数据也
discrepant the robot’s progress was from the optimal solution to entrap the moving target. 提供了足够的信息。同样，个体差异很重要。在pretest上具有较高空间能力得分的参与
Overall, 86% of the participants successfully entrapped the target using the full automation 者具有更好的目标获取率，而经验丰富的游戏玩家比经验不足的同龄人更擅长封装移动目
solution, whereas 96% of the participants did so with the hybrid solution (without visualization), 标（Chen，2011;2012）。
suggesting the possible advantages of human/autonomy collaboration found in the University of
Central Florida (UCF) studies discussed in section 4. Visualization aiding had little impact on
performance; even with partial autonomy, the raw data on the map display supplied sufficient
information. Again, individual differences were important. Participants with higher spatial ability
scores on the pretest had better target acquisition rates, whereas experienced gamers were better
at encapsulating the moving target than their less experienced peers (Chen, 2011; 2012).
The design implications are as follows: 设计含义如下:

• Intelligent agents acting as surrogate intermediate supervisors are a potentially effective • 充当代理中间监督员的智能代理是控制多个自治系统的潜在有效方法。
way of controlling multiple autonomous systems.
• At a minimum, agent/human teams must have two characteristics: operators must have • 至少，代理/人工团队必须具有两个特征：操作员必须具有最终决策权，代理必
final decision authority and agents must signal their intentions clearly. 须清楚地表达他们的意图。
• Results in the Chen and Barnes (2012b) experiment suggest that for agents with moderate • Chen和Barnes(2012b)实验中的结果表明，对于具有中等错误率（约80％）的代理
error rates (approximately 80%), FAP (vs. MP) weighted alerts can be a relatively efficient ，如果FA很容易检查，fap（与MP）加权警报可以是提醒潜在问题的相对有效手段
means of alerting potential problems if the FA are easily checked. 。
• Individual differences in spatial abilities, attentional control, and gaming experience are • 空间能力，注意力控制和游戏体验的个体差异是人类与监管多种资产的代理人互
important determinants of how well humans interact with agents supervising multiple 动程度的重要决定因素。
assets.
5 5
4. Safety and Levels of Autonomy 4. 安全和自主水平
As mentioned in the introduction, there were two main thrusts for the SOURCE program: 正如引言中提到的，源程序有两个主要的推动力：增强自主能力和确保人类在自主系统
enhancing autonomous capabilities and ensuring safety for humans in the vicinity of autonomous 附近的安全。 UCF研究人员专注于各种LOA的影响以及它们如何影响任务绩效，操作员工
systems. UCF researchers focused on the effects of various LOA and how they influenced 作量，信任，SA以及最重要的是它们如何影响人类安全。最初的实验是在一个微型城市
mission performance, operator workload, trust, SA, and, most important, how they affected 环境（1/35比例）中进行的，代表典型的伊拉克城市地区的六个正方形街区（图2）。测
human safety. The initial experiments were conducted in a miniature urban setting (1/35th scale) 试参与者被安置在一个与世隔绝的房间里，有多个电视监视器，让他们可以看到正在进行
的实验。他们不知道实验者在一个单独的房间里，实际上根据参与者的输入对小型化的U
representing six square blocks of a typical Iraqi urban area (figure 2). The test participants were
v进行了远程操作(Fincannon et al., 2011 ).
stationed in an isolated room with multiple television monitors that allowed them to view the
ongoing experiments. They were unaware that the experimenters, in a separate room, actually
teleoperated the miniaturized UVs based on the participants’ inputs (Fincannon et al., 2011).
Figure 2. Views of Middle Eastern urban setting as viewed by operators in the control room. 图2。控制室内操作员所看到的中东城市环境景观。
The first two experiments investigated robot-to-robot autonomy (Jentsch et al., 2010; Phillips et 前两个实验研究了机器人对机器人的自主性(Jentsch et al.，2010;Phillips等人。, 2010).
al., 2010). The task was to coordinate four UVs to transverse one of five urban routes and engage 任务是协调四条Uv横贯五条城市路线中的一条，并根据Uv之间的通信流量找到最佳路线
a predetermined target by finding the optimal route based on communication traffic among the ，从而达到预定目标。 Jentsch等人。（2010）将自主条件（机器人可以自动协调）与手
UVs. Jentsch et al. (2010) compared autonomy conditions (robots could automatically 动条件（人类操作员查看通信信息但做出协调决定）进行了比较。在这两种情况下，人
coordinate) with manual conditions (human operators viewed the communications information 类都做出了最终决定。在前一种情况下，这是一个被动的决定，通过肯定或否认自动决
定，而在后一种情况下，运营商积极参与协调Uv。 Uv之间的通信是100％或80％可靠的
but made the coordination decisions). In both cases, the human made the final decision. In the
。具体来说，80%的可靠性是通过在通信过程中故意添加不正确的信息来实现的。由于
former case it was a passive decision by affirming or denying the automated decisions, whereas
参与者没有时间或路线限制，协调任务只有中等难度。对于不完全可靠的条件，操作员
in the latter case operators were actively involved in coordinating UVs. Communication among 在自己做出协调决定时表现更好。因此，对于80%的可靠性条件，让操作员积极参与决
the UVs was either 100% or 80% reliable. Specifically, 80% reliability was accomplished by 策过程改善了协调性能。
deliberately adding incorrect information during communications. Because participants did not
have time or route constraints, the coordination task was only moderately difficult. For the less
than perfectly reliable conditions, operators performed better when they made the coordination
decision themselves. Thus, for the 80% reliability condition, having the operator actively
involved in the decision process improved coordination performance.
6 6
In the second experiment task difficulty was increased by adding “no-go” areas and time limits. 在第二个实验中，增加了"禁止进入"区域和时间限制，增加了任务难度。
Under these conditions, autonomy was superior to manual operator coordination in both 在这些条件下，在两种可靠性条件下，自主性都优于手动操作员协调。这一发现突出了
reliability conditions. This finding highlights the importance of the interaction between task 任务难度和自主性之间相互作用的重要性，以及将人类保持在循环中的限制。因此，为
difficulty and autonomy, as well as the limits of keeping the human in the loop. Thus, to 了优化性能，系统设计者需要找到让人处于决策循环和依赖自治系统之间的"最佳点"，
optimize performance, the system designer needs to find the “sweet spot” between keeping the 特别是当任务难度增加到超出人有效执行任务的能力时。这与自适应自治的研究结果类
human in the decision loop and dependence on autonomous systems, particularly as task 似，表明当任务负载可管理时，让人类参与手动控制并在更困难的情况下调用自治所获
得的效率（参见2007年12月17日）。巴恩斯等人。，2006;Parasuraman等人。, 2007).
difficulty increases beyond the human’s ability to perform the task effectively. This is similar to
findings in adaptive autonomy showing the efficiencies gained in having the human engage in
manual control when task load is manageable and invoking autonomy under more difficult
situations (cf. Barnes et al., 2006; Parasuraman et al., 2007).
A more recent UCF experiment varied automation and degree of human involvement during 最近的UCF实验在1/35规模伊拉克设施的模拟过程中改变了自动化和人类参与程度(Finca
simulations in the 1/35th scale Iraqi facility (Fincannon et al., 2012). For this study, the nnon et al., 2012). 在这项研究中，研究人员将机器人的任务分解为三个部分：（1）检测
researchers decomposed the robots’ tasks into three components: (1) detect a possible significant 一个可能的重要物体，并决定停止机器人以大约5英里/小时的模拟速度行驶，（2）识别
object and make a decision to stop the robot traveling at simulated speeds of approximately 5 物体的类型，以及（3）根据当前的参与规则（ROEs）决定要采取的行动类型。 ROEs是
mph, (2) identify the type of object, and (3) decide the type of action to be taken based on the 命令发布的规则，允许士兵在允许的指导方针下执行任务。在实验中，Roe由研究人员开
发，并在每次会议之前给予参与者。在手动条件下，所有任务都由人工操作员执行，而
current rules of engagement (ROEs). ROEs are command-issued rules that permit Soldiers to
在自主条件下，即使所有任务都是自动化的，操作员也可以选择复盖任务2和3的自主性
conduct their missions under permissible guidelines. In the experiment, the ROEs were
。最后，在协作条件下，任务2由操作员执行，任务1和3被自动化。协作条件利用了自
developed by the researchers and given to the participants before each session. In the manual 主性和代理感知的障碍检测优势-即人类操作员对目标识别的感知优势。
condition, all tasks were performed by the human operator, whereas in the autonomy condition,
even though all tasks were automated, operators were given the option of overriding the
autonomy for tasks 2 and 3. Finally, in the collaborative condition, task 2 was performed by the
operator and tasks 1 and 3 were automated. The collaborative condition took advantage of both
the obstacle detection strengths of autonomy and perception by proxy—that is, the human
operator’s perceptual strengths for target identification.
The most dramatic differences were evinced in task-1 detection of possible targets and stopping 最显着的差异表现在task-1检测可能的目标和停止机器人：手动准确率为37％，自主和
the robot: 37% accuracy for manual, and 67% and 58% accuracy for the autonomy and 协作条件的准确率分别为67％和58％。这表明，手动控制机器人的操作员发现很难发现
collaborative conditions, respectively. This indicated that an operator controlling a robot 并对意外事件做出反应，即使是不完美的自动化系统也比仅依靠操作员完成这项任务更安
manually found it very difficult to spot and react to unexpected events, and that even imperfectly 全。然而，对于综合信息（任务2和3），人工和智能系统的组合（协作）决策优于自主
automated systems are safer than relying solely on the operator for this task. However, for 或手动控制条件，除非操作员的工作量很高。因此，结果表明，自主性可以通过检测机
器人路径中的重要物体来增强安全性，但人类也可以通过能够识别这些物体（通过代理感
synthesizing information (tasks 2 and 3), a combined (collaborative) human and intelligent
知）来发挥重要作用。因此，人类参与是有用的警告，当工作量太高时，人类可能会变
system decision was superior to either autonomous or manual control conditions except when the
得不堪重负。
operator’s workload was high. Thus, the results suggest that autonomy can enhance safety by
detecting significant objects in the robot’s path, but that humans can also play an important role
by being able to identify these objects (perception by proxy). Therefore, human involvement is
useful with the caveat that humans can become overwhelmed when the workload is too high.
The final experiment investigated LOAs using the MIX test bed similar to the RoboLeader 最后的实验使用类似于前面描述的RoboLeader研究的混合试验台研究了LOAs。与1/35
studies described previously. Compared to the 1/35th scale world, the MIX computer simulation 比例的世界相比，混合计算机模拟环境可以更精确地控制模拟参数，如车辆速度和行人
environment allowed for more precise control of simulation parameters, such as vehicular speed 过路(Jentsch等)。，2011;Sellers等人。, 2012). 再次强调的是
and pedestrian crossings (Jentsch et al., 2011; Sellers et al., 2012). Again, the emphasis was on
7 7
safety and LOA; however, the objective was to investigate the effectiveness of two operator 安全和LOA;然而，目标是调查两种操作员干预策略的有效性。参与者被告知，行人将在
intervention strategies. The participants were told that pedestrians would transverse the robot’s 三种LOA条件之一下横向机器人的路径：（1）完全自主（AU），（2）同意管理（MBC
path under one of three LOA conditions: (1) fully autonomous (AU), (2) management by consent ）和（3）例外管理（MBE）。非盟系统根据ROEs选择了一个响应（例如，继续情报表明
(MBC), and (3) management by exception (MBE). The AU system chose a response based on 存在危险情况），而ROEs反过来又基于每个模拟小插曲的封面故事。对于MBC条件，自
the ROEs (e.g., continue—intelligence suggests a dangerous situation), which were, in turn, 治将始终停止机器人并建议行动方案;操作员必须同意或更改AU决定才能继续。相比之下
，在MBE条件下，参与者可以复盖自主决策。在参与者没有选择推翻自主决定的情况下
based on the cover story for each simulated vignette. For the MBC condition, the autonomy
，将执行自主选择的行动方针。实验者还改变了可靠性（60％正确或90％正确）。正确
would always stop the robot and suggest a course of action; the operator had to consent to or
性是基于非盟对每个小插曲给予操作员的ROEs的遵循程度。
change the AU decision before continuing. In contrast, in the MBE condition, participants could
override the autonomous decision. In situations where participants did not choose to override the
autonomous decision, the autonomy-chosen course of action would be executed. The
experimenters also varied autonomy reliability (either 60% correct or 90% correct). Correctness
was based on how well the AU followed the ROEs given to the operators for each vignette.
Overall, operators in the MBE conditions showed significantly superior performance (executing 总体而言，与AU和MBC相比，MBE条件下的操作员表现出显着优越的性能（执行正确的
the correct ROE for safety) when compared to both AU and MBC. Figure 3 shows a significant ROE以确保安全）。图3显示了可靠性和LOA之间的显着相互作用。 MBE条件允许操作员
interaction between reliability and LOA. The MBE condition allowed operators to take advantage 在高可靠性条件下利用AU条件的准确性，但也导致操作员在低可靠性任务段中推翻较差
of the accuracy of the AU condition in high reliability conditions but also resulted in the operator 的AU决策。 MBC运营商表现出更大的倾向，错误地进行二次猜测高度可靠的自治。这种
overriding poor AU decisions during low-reliability mission segments. MBC operators showed a 解释是由这样一个事实支持的，即运营商倾向于给MBC条件比全自动条件更高的信任分数
。因此，对于MBC来说，与高度可靠的自主决策相比，运营商过度信任自己的决策。这
greater tendency to incorrectly second-guess highly reliable autonomy. This interpretation is
可能部分是因为与MBE相比，他们需要额外的时间来做出决策，这使得运营商更有可能猜
buttressed by the fact that operators tended to give higher trust scores to MBC conditions than to
测甚至高度可靠的自治(Beck et al., 2007). 虽然Jentsch及其同事的研究很复杂，但他们表
fully automated conditions. Thus for MBC, operators over-trusted their own decision compared to 明，只有当操作员的整体任务绩效水平不允许操作员停留在决策循环中时，完全自主才是
highly reliable autonomous ones. This could be in part because of the additional time they had to 理想的。这些研究的结果还表明，信任本身并不是HRI的目标;相反，信任应该进行适当的
make decisions compared to MBE, making the operator more likely to second-guess even highly 校准-在设计自主系统的人类校准时，知道何时信任和何时复盖是目标(Chen et al.，2007;
reliable autonomy (Beck et al., 2007). Although the studies by Jentsch and colleagues are complex, Lee和See，2004）。特别是在战斗中，人类将永远是安全问题的最终仲裁者，因此，他
they suggest that full autonomy is only ideal when the operator’s overall tasking performance level 们必须知道自治何时比人类干预更安全，反之亦然（Barnes和Evans，2010;Chen和Barne
does not permit the operator to stay in the decision loop. The findings of these studies also suggest s，正在审查中;也见Goodrich和Blatter，2012）。
that trust per se is not the objective of HRI; instead, trust should be appropriately calibrated—
knowing when to trust and when to override is the objective when designing for human calibration
of autonomous systems (Chen et al., 2007; Lee and See, 2004). Especially in combat, humans will
always be the final arbitrator of safety issues, and, as such, they must know when autonomy is
safer than human intervention and vice-versa (Barnes and Evans, 2010; Chen and Barnes, under
review; also see Goodrich and Blatter, 2012).
Figure 3. Interaction between reliability and LOA. 图3。可靠性和LOA之间的相互作用。
8 8
• The degree of human decision involvement in autonomous systems depends on two factors: • 自主系统中人为决策的参与程度取决于两个因素：自动化的有效性和人类操作员的
the effectiveness of the automation and the human operator’s tasking limitations. For 任务限制。例如，当操作员完全参与多任务时，即使是不完美的自动化也可能优先
example, even imperfect automation may be preferred to human involvement when the 于人类参与。
operator is fully engaged in multitasking.
• Adaptive automation (AA) wherein the human is re-engaged in the decision process for • 自适应自动化（AA），其中人类在可管理的任务段期间重新参与自动化任务的决
automated tasks during manageable mission segments is preferred to full automation. This 策过程是首选的完全自动化。这是因为AA让操作员知道自动化任务的状态，但只
is because AA keeps the operators aware of the state of the automated task but only 有在总体多任务级别可管理时才会参与它们。
engages them when the overall multitasking level is manageable.
• Autonomy can improve robotic safety by being able to respond to potentially dangerous • 自治可以通过能够比人类更快地响应潜在危险情况来提高机器人的安全性，例如在
situations more rapidly than humans can, such as the sudden appearance of a pedestrian in 复杂的城市环境中突然出现行人。
complex urban environments.
• A possible strategy for overcoming autonomy limitations is developing hybrid systems that • 克服自主性限制的一个可能的策略是开发混合系统，允许人类做他们最擅长的事情
allow humans to do what they do best, such as interpreting the significance of detected ，例如解释检测到的物体的意义（通过代理感知）。
objects (perception by proxy).
• Overall, in the UCF studies, the MBE LOA that allowed humans to override autonomous • 总体而言，在UCF研究中，与AU和MBC相比，允许人类推翻自主决策的MBE LOA
decisions was the most effective strategy compared to AU and MBC. MBE resulted in 是最有效的策略。与低可靠性自治相比，MBE的决策更安全，但与高可靠性自治
safer decisions than did low reliability autonomy, but MBE showed a minimal loss of 相比，MBE的决策准确性损失最小。
decision accuracy when compared to highly reliable autonomy.
• Trust, as measured by the UCF subjective scale, was a poor predictor of performance. For • 信任，由UCF主观量表衡量，是绩效的一个较差的预测指标。例如，mbc条件下的
example, trust was higher in the MBC condition than the AU condition. However, overall 信任度高于AU条件。然而，在MBC任务阶段，总体人力表现要比MBE或AU条件差
human performance was poorer during MBC mission segments than for either the MBE or 。
AU conditions.
5. Ft. Benning and Ft. Leonard Wood Field Experiments: Progressive 5. 英尺。本宁和英国《金融时报》。伦纳德*伍德野外实验：进步
Autonomy and Intuitive Interfaces 自主性和直观的界面
This section covers issues related to autonomy and intuitive interfaces for small robots. The 本节涵盖与小型机器人的自主性和直观界面相关的问题。研究人员在使用现役士兵作为
researchers investigated a variety of interfaces for SUGV applications during exercises in field 测试参与者的野外环境中进行演习期间调查了SUGV应用的各种接口。这些接口要么加快
environments using active duty Soldiers as test participants. These interfaces either expedited 了语音控制等监控功能，要么增强了操作员远程执行SUGV任务的能力（远程呈现和立体
supervisory functions, such as voice control, or enhanced the operator’s ability to conduct SUGV 视觉）。探讨了不同程度的小型机器人自主化的优缺点。
missions remotely (telepresence and stereovision). The advantages and disadvantages of
different degrees of small robot autonomy were explored as well.
HRED researchers worked with Navy researchers from the Space and Naval Warfare Systems HRED研究人员与加利福尼亚州圣地亚哥的太空和海战系统中心（SPAWAR-Pacific）
Center (SPAWAR-Pacific) in San Diego, CA, to explore progressive autonomy for SUGVs in 的海军研究人员合作，探索SUGVs在
9 9
field experiments at Ft. Benning. Full autonomy entails being able to respond to the environment 在Ft现场实验。本宁。完全自主意味着能够在意想不到的情况下对环境做出反应，并从
in unexpected circumstances and to learn from past experience (Pettitt et al., 2010; Pettitt et al., 过去的经验中学习(Pettitt et al.，2010;Pettitt等人。, 2012). 渐进式自治假设自治将以循
2012). Progressive autonomy assumes that autonomy will progress from manual control to 序渐进的方式从手动控制进展到不同程度的自治。陆军和海军研究人员团队研究了三个
different degrees of autonomy in a step-wise fashion. The team of Army and Navy researchers 层次：远程操作，部分自治和完全自治（具有可预测的环境并且没有机器学习）。因此
examined three levels: teleoperations, partial autonomy, and full autonomy (with a predictable ，即使是完全自治条件也被简化为完全自治的严格定义。实验的目的是研究随着控制逐
渐变得更加自主，对任务和人类表现的影响。远程操作要求操作员在执行辅助任务（在
environment and without machine learning). Thus, even the full autonomy condition was
某些情况下）时手动控制产卵机器人（图4）。部分自主包括在途中自主避障；然而，除
simplified in terms of a strict definition of full autonomy. The purpose of the experiment was to
非部分自主超越操作员以避开障碍物，否则操作员控制机器人的运动。全自主软件自动
investigate the effects on mission and human performance as control became progressively more 执行机器人移动到航点和避障.
autonomous. Teleoperations required the operator to manually control the SPAWAR robot
(figure 4) while performing secondary tasks (for some conditions). Partial autonomy consisted of
autonomous obstacle avoidance in-route; however, the operator controlled the movement of the
robot unless the partial autonomy overrode the operator to avoid obstacles. Full autonomy
software automatically performed both robot movement to waypoints and obstacle avoidance.
Figure 4. SPAWAR robot. 图4。 SPAWAR机器人.
In general, full autonomy resulted in zero driving errors (superior to the partial condition which, 一般来说，完全自主导致零驾驶错误（优于部分条件，而部分条件的错误又比手动条件
in turn, had significantly fewer errors than then the manual condition), faster course times, and 少得多），更快的课程时间，以及比部分条件和手动条件更少的停止执行次要任务。部
fewer stops to conduct secondary tasks than both the partial and manual conditions. Partial 分自治比手动和完全自治条件都要慢得多，也被评为更令人沮丧。识别潜在目标和次要
autonomy was significantly slower than both the manual and full autonomy conditions and was 任务主要影响的能力不显着。次要任务影响了手动和部分自治系统的任务时间，但对完
also rated as more frustrating. The ability to identify potential targets and secondary task main 全自治没有影响。自治系统的优越性是可以预料的，但它确实表明，即使对于SUGV来
说，某种程度的自治也正在迅速成为最先进的技术。对于部分自主条件的挫败感是可以
effects was not significant. Secondary tasks affected mission time for manual and partially
理解的，因为在部分自主条件下，操作员报告有时不得不"对抗"软件以找到绕过障碍物
autonomous systems but had no effect on full autonomy. The superiority of the autonomous
的正确方法。为
systems was to be expected, but it does indicate that some degree of autonomy is rapidly
becoming state of the art even for the SUGV. The frustration for the partially autonomous
conditions was understandable because in the partially autonomous condition the operators
reported sometimes having to “fight” the software to find the correct way around obstacles. For
10 10
the same reasons, mission time was significantly longer for the partial condition than for either of 同样的原因，部分条件的任务时间比其他两个实验条件中的任何一个要长得多。然而，
the other two experimental conditions. However, even partial autonomy was superior to the 即使是部分自治也优于手动条件，以减少驾驶错误并防止不必要的停车来执行次要任务。
manual condition for reducing driving errors and preventing unnecessary stops to conduct 因此，部分自治显示出一些优势，但总的来说，它不如完全自治有利，因为部分自治可
secondary tasks. Thus partial autonomy showed some advantages, but in general, it was less 能令人沮丧并导致反应时间增加。
advantageous than full autonomy because partial autonomy can be frustrating and result in
increased reaction times.
In another series of experiments, ARL researchers at Ft. Benning working with a private 在另一系列实验中，Ft的ARL研究人员。 Benning与一家私人公司（Think-A-Move*）合
company (Think-A-Move *) to evaluate voice interfaces during realistic dismounted field 作，在现实的下马现场实验中评估语音接口。他们的研究目标是评估先进的界面设计，
experiments. Their research objectives were to evaluate advanced interface designs to improve 以改善SA，以及释放士兵的手和眼睛的正面行动。机器人资产的语音控制具有许多明显
SA as well as free the Soldier’s hands and eyes for head’s up operations. Speech control of 的优势：（1）士兵与机器人互动的自然方式，这反过来促进了团队关系，（2）随着SUG
robotic assets has a number of distinct advantages: (1) it is a natural way for Soldiers to interact Vs变得更加自主，它具有免提控制的潜力。 HRED研究进行了许多研究，评估了语音对小
型机器人控制的有效性(Petitt et al.，2012;Redden等人。, 2013). 他们的目标是表明语音
with robots, which in turn fosters a team relationship, and (2) it has the potential of hands- and
控制可以通过用更轻，更小的语音系统替换手动控制器来减小控制器的尺寸（图4）。实
eyes-free control as SUGVs become more autonomous. The HRED researches conducted a
验是使用遥控机器人进行的，但结果可以在半自主机器人和控制杂项功能（如菜单选择）
number of studies evaluating the efficacy of voice for small robot control (Petitt et al., 2012; 所需时转移到操作员干预。他们发现，基于语音的控制在控制器尺寸缩小之外还具有性
Redden et al., 2013). Their goal was to show that speech control could reduce the size of the 能优势。具体来说，它通过允许语音命令来控制机器人手臂，同时使用手动控制来操纵
controller by replacing the manual controller with a lighter, smaller speech system (figure 4). 机器人，从而减少了同时执行多个任务的时间和精力。然而，如果必须控制机械臂的平
The experiments were conducted using teleoperated robots, but the results could transfer to 移和倾斜，士兵们在基于语音的控制方面遇到了麻烦，因为语音命令是离散的，并且缺乏
operator interventions when necessary for semi-autonomous robots and for controlling 手动控制可能的精确运动。此外，ARL研究人员调查了士兵被要求执行的各种任务的直观
miscellaneous functions, such as menu selection. They found that speech-based control exhibited 词汇，使研究人员能够为实验实施以用户为中心的词汇。
performance benefits beyond controller size reduction. Specifically, it decreased time and effort
when performing multiple tasks simultaneously by allowing speech commands to be given for
control of the robotic arm while at the same time maneuvering the robot using manual controls.
However, the Soldiers had trouble with the speech-based control if they had to control the pan
and tilt of the robotic arm because the voice commands were discrete and lacked the precise
movements possible with manual control. Also, the ARL researchers investigated intuitive
vocabularies for the various tasks that Soldiers were asked to perform, allowing the researchers
to implement a user-centered lexicon for the experiment.
In the second speech experiment, Redden et al. (2013) investigated the potential for using speech 在第二个语音实验中，Redden等人。 (2013)研究了将语音用于多用途功能的可能性，例
for multipurpose functions, such as having the robot photograph improvised explosive devices 如让机器人拍摄简易爆炸装置（Ied）或让操作员选择菜单上的项目。当操作员被要求执
(IEDs) or having the operator choose items on a menu. When the operator was required to 行次要任务时，语音控制提高了多任务性能，因为语音对于共享认知的效率。同样，当
perform a secondary task, speech control improved multitasking performance because of the 操作员必须访问菜单才能拍摄潜在的IED（例如，"放大图片"）时，语音控制明显快于手
efficiency of speech for shared cognition. Similarly, when the operator had to access a menu in 动控制。然而，通过操纵机器人拍摄照片使用手动控制更有效，因为操纵机器人是一个
连续的过程。
order to take picture of a potential IED (e.g., “enlarge a picture”), speech control was
significantly faster than manual control. However, taking a photo by maneuvering the robot was
more efficient using manual control because maneuvering the robot is a continuous process.
In a different domain, Elliott and her Ft. Benning colleagues (Elliott et al., 2012; Jansen et al., 在另一个领域，艾略特和她的Ft。本宁同事（Elliott et al.，2012;Jansen等人。，201
2012; Redden et al., 2013) collaborated with researchers from the TNO laboratories in the 2;Redden等人。（2013年）与TNO实验室的研究人员合作。
*Think-A-Move is a trademark of Think-A-Move Ltd. *Think-A-Move是Think-A-Move Ltd.的商标。
11 11
Netherlands to evaluate telepresence techniques designed to artificially create the feeling of 需要评估旨在人为地创造在机器人正在观看的区域内的感觉的临场感技术。远程呈现的
being in the area the robot is viewing. The obvious advantage of telepresence is that an 明显优势是，自主机器人能够收集感兴趣的指定区域（NAO）的信息，而不会使士兵受
autonomous robot would be able to gather information for a named area of interest (NAO) 到伤害。增加机器人的视频尤为重要，因为以前的研究表明，机器人的视频馈送给出了n
without putting the Soldier in harm’s way. Augmenting a robot’s video is particularly important ao的贫困视图(Chen et al., 2007). 临场感增强包括立体视觉和头戴式摄像机，操作员可以
because previous research indicated that video feed from robots gives an impoverished view of 用它来扫描远程区域，模拟实际在该区域。在第一个实验中，任务相对容易，目标检测
，SA或工作量测量差异与常规接口没有显着差异。然而，士兵参与者更喜欢临场感，这
the NAO (Chen et al., 2007). The telepresence augmentations included stereovision and a head-
表明进一步的研究可能是值得的。第二个临场感实验包含更困难的检测任务，临场感通
mounted camera that the operator could use to scan the remote area emulating actually being in
过使用三维（3-D）音频线索来定位目标进一步增强(Elliott et al., 2012). 将远程呈现与头
the area. In the first experiment, the tasking was relatively easy, and target detection, SA, or 盔式显示器（HMD）和操纵杆进行比较，以在远程位置定位目标。此外，还有与每个目
workload measurement differences were not significantly different from conventional interfaces. 标相关联的声音。与HMD和操纵杆相比，3-D音频增强提高了工作负载减少、响应速度
However, Soldier participants preferred telepresence, suggesting further research might be 和目标识别的性能。在消极方面，临场感设备体积庞大，不适合步兵行动。
worthwhile. The second telepresence experiment contained more difficult detection tasks, and
the telepresence was augmented further by using three-dimensional (3-D) audio cues to locate
targets (Elliott et al., 2012). Telepresence was compared to a helmet-mounted display (HMD)
and a joystick to locate targets in a remote location. In addition, there was sound associated with
each target. The 3-D audio augmentation resulted in improved performance compared to the
HMD and joystick for workload reduction, speed of responses, and target identification. On the
negative side, the telepresence equipment was bulky and not ideally suited for infantry
operations.
Another important issue related to SUGVs is their use for finding and defeating IEDs. IEDs have 与SUGVs相关的另一个重要问题是它们用于寻找和击败Ied。简易爆炸装置已被证明是特
proved to be particularly deadly and difficult to detect. Progressive autonomy would make this 别致命和难以检测。渐进式自治将使这项任务更容易，但最重要的问题涉及远程操纵IED
task easier, but the most important issue involves manipulating the IED remotely to disarm it. 以解除其武装。目前，用于与Ied一起工作的机器人具有设备的二维（2-D）视图，使得
Currently, the robots used to work with IEDs have a two-dimensional (2-D) view of the device, 任务比手动解除IED更困难，因为操作员缺乏与正常人类视觉相关的3-D线索（Chen et al.,
making the task more difficult than manually disarming an IED because the operator lacks the 2007). 许多先前的研究已经确定了立体视觉在相关任务期间用于SUGV显示器的有用性（
Barnes和Evans，2010）。最近在英国《金融时报》进行的实地实验。 Leonard Wood研
3-D cues associated with normal human vision (Chen et al., 2007). A number of previous studies
究了立体视觉显示器和触觉反馈的组合，用于IED操作(Edmondson et al., 2012). 北极星传
have identified the usefulness of stereovision for SUGV displays during related tasks (Barnes and
感器技术和哈里斯公司与ARL和非盈利机构合作
Evans, 2010). More recent field experiments at Ft. Leonard Wood examined the combination of
stereovision displays and haptic feedback for IED manipulations (Edmondson et al., 2012).
Polaris Sensor Technologies and Harris Corporations working with ARL and the nonprofit
Leonard Wood Institute evaluated an interface suite to improve Soldier safety using a Talon * 伦纳德伍德研究所评估了一个接口套件，以提高士兵的安全使用爪*机器人寻找，操纵和
robot to find, manipulate, and disarm IEDs. The current study incorporated not only a stereovision 解除简易爆炸装置。目前的研究不仅包括立体视觉显示器，还包括Harris控制器†，可向
display but also a Harris controller† that gave haptic feedback to the operator (figure 5). 操作员提供触觉反馈（图5）。
*Talon is a registered trademark of QuintQ_North Amarica. *Talon是QuintQ_North Amarica的注册商标。 †哈里斯控

†Harris controller is a registered trademark of Harris Corporation. 制器是哈里斯公司的注册商标。
12 12
Figure 5. The haptic manipulator used with stereovision for the Talon 图5。触觉机械手与立体视觉一起用于爪机器人实验.
robot experiment.
Nine participants performed navigation, search, and arm manipulation tasks for scenarios that 九名参与者执行导航，搜索和手臂操纵任务，用于指示美国陆军工程，宪兵和生化任务
were indicative of U.S. Army engineering, military police, and biochemical missions. There 的场景。有统计学显着的延迟效应（3-D与2-D）和控制器条件有利于3-d触觉组合的
were statistically significant latency effects of view (3-D vs. 2-D) and insignificant trends for 微不足道的趋势。同样，这两种情况对预期的工作量减少也有显着影响。与会者还支持
controller conditions favoring the 3-D haptic combination. Similarly, there were significant 触觉和立体视觉组件单独和作为一个组合单元。总之，结果表明用户接受以及立体视觉
effects for both these conditions for perceived workload reductions. The participants also 和触觉控制器的性能改进，特别是当组合在同一界面时。
endorsed haptic and stereovision components individually and as a combined unit. In summary,
the results indicated user acceptance as well as performance improvements for stereovision and
haptic controllers especially when combined in the same interface.

• Progressive autonomy research indicates that SUGVs with obstacle avoidance and • 渐进式自主研究表明，具有避障和自主移动到航点的SUGVs可以比远程操作的SUGV
autonomous movement to waypoints can perform simple search tasks more effectively than s更有效地执行简单的搜索任务。
teleoperated SUGVs.
• Partial autonomy, such as obstacle avoidance, can introduce problems as well as increase • 部分自治，例如避障，如果与运营商的控制策略相冲突，可能会带来问题并提高
effectiveness if they conflict with operators’ control strategies. 效率。
• It is important to tailor speech commands to the target audience. Tailoring allows better • 为目标受众定制语音命令非常重要。剪裁允许更好的保留和更有效的操作。
retention and more efficient operation.
• Because of its lower cognitive demand, speech control is more rapid than manual control in • 由于其较低的认知需求，语音控制在需要多任务处理的情况下比手动控制更快速。
situations that require multitasking.
• Manual control is more effective than speech control for nondiscrete tasks (e.g., turning); • 对于非分析任务（例如，转弯），手动控制比语音控制更有效;然而，语音对于菜
however, speech is more efficient for menu selection. 单选择更有效。
13 13
• Although not currently configured for efficient infantry uses, telepresence has great • 虽然目前没有配置为高效的步兵使用，但远程呈现在使用机器人资产的战斗
potential for remote sensing of combat environments using robotic assets. 环境遥感方面具有很大的潜力。
• A combination of haptic feedback and stereovision shows promise for safely manipulating • 触觉反馈和立体视觉的结合显示了使用小型机器人安全操纵和拆除爆炸和化学装置
and defusing explosive and chemical devices using small robots. 的前景。
6. Situation Understanding and Decision Support 6. 情况理解和决策支持
Although we can expect autonomous systems to become more prevalent as the technology 虽然我们可以预期随着技术的进步，自主系统将变得更加普遍，但在可预见的未来，某
improves, certain functions such as situation understanding will remain predominantly human 些功能（如情况理解）将仍然主要是人类功能（Barnes和Evans，2010）。这是因为情
functions for the foreseeable future (Barnes and Evans, 2010). This is because situation 况理解不仅涉及当前的SA，还涉及道德问题和对不断变化的情况的长期影响的考虑（Ch
understanding entails more than current SA but also involves ethical issues and consideration of en和Barnes，in press）。对于情报收集来说尤其如此，这些情报收集可能涉及微妙的线
the longer-term implications of the evolving situation (Chen and Barnes, in press). This is 索和政治细微差别，这些细微差别每天都在变化。来自以色列本古里安大学（BGU）的
Oron-Gilad和她的同事进行了一系列实验，以提高士兵从多个无人系统收集情报信息的
particularly true for intelligence gathering which may involve subtle cues and political nuances
能力（Oron-Gilad et al.，2011;Oron-Gilad，在准备中）。
that change from day to day. Oron-Gilad and her associates from Ben-Gurion University (BGU)
in Israel conducted a series of experiments to enhance the Soldier’s ability to glean intelligence
information from multiple unmanned systems (Oron-Gilad et al., 2011; Oron-Gilad, in
preparation).
Oron-Gilad et al.’s (2011) initial experiments evaluated the utility of a physical device to portray Oron-Gilad等人。's（2011）的初步实验评估了物理设备向下马的以色列国防军（IDF）
the unmanned vehicle’s (UV’s) information to the dismounted Israel Defense Forces (IDF). The 描绘无人驾驶车辆（UV）信息的效用。显示器包含地图和视频部分。结果取决于地形类
displays contained both a map and a video portion. The results depended on both the terrain type 型和任务。具体而言，对于细粒度分析，需要更大的显示器（视频部分：对角线为7.5）
and the task. Specifically, for fine-grained analysis a larger display was required (video portion: ，而对于其他任务，例如在农村环境中操作，视频部分对角线为3.5的显示器就足够了。
7.5-in diagonal), whereas for other tasks such as operations in a rural environment, a display with Oron-Gilad等人。与HRED人员合作，在Ft使用美国现役军人进行了随后的实验。本宁。
他们检查了三种显示器类型：12英寸平板电脑（7.5英寸视频对角线），手持显示器（HH
a video portion of 3.5-in diagonal was sufficient. Oron-Gilad et al., working with HRED
D，4.2英寸视频对角线）和单眼头盔式显示器（HMD，3.2英寸）。在这个实验中，平板
personnel, conducted a subsequent experiment using U.S. active duty Soldiers at Ft. Benning.
电脑和HHD在报告情报指标（人员，车辆和运动）方面没有差异。然而，HMD的性能明
They examined three display types: a 12-in tablet (7.5-in video diagonal), a handheld display 显低于其他两款器件。参与者还报告了HMD应用的双眼竞争和眼睛疲劳。研究人员还研
(HHD, 4.2-in video diagonal), and a monocular helmet-mounted display (HMD, 3.2 in). In this 究了UAS与UGV图像在情报收集任务中的相对有效性。有些令人惊讶的是，他们发现自我
experiment, there was no difference between the tablet and HHD for reporting intelligence 中心观点（UGV）优于外心UAS条件。这可能是情景特定的，因为其他实验表明相反的效
indicators (personnel, vehicles, and movement). However, the HMD resulted in significantly 果（Barnes和Evans，2010）。
poorer performance than the other two devices. Participants also reported binocular rivalry and
eyestrain for the HMD applications. The researchers also examined the relative effectiveness of
UAS vs. UGV imagery for the intelligence-gathering task. Somewhat surprisingly, they found
that the egocentric view (UGV) was superior to the exocentric UAS conditions. This may be
scenario specific because other experiments indicated the opposite effect (Barnes and Evans,
2010).
In two experiments, Ophir-Arbelle et al. (2013) evaluated interfaces that showed both UAS and 在两个实验中，Ophir-Arbelle等人。 (2013)评估了同时显示UAS和
UGV imagery. These interfaces displayed multiple views of the area of operations. Because they UGV图像。这些界面显示了操作区域的多个视图。因为他们
14 14
assumed that UAS would be the default mode for IDF operations, they examined two conditions: 假设UAS是IDF操作的默认模式，他们检查了两个条件：单独的UAS和组合的UAS/UGV（
UAS alone and combined UAS/UGV (figure 6). The video for both UVs was urban feeds with 图6）。这两个UVs的视频都是与士兵或车辆移动有关的目标的城市饲料。任务是确定随
targets related to Soldiers or vehicular movement. The tasks were to identify the randomly 机出现的目标，并执行一个需要协调视频运动与地图方向的方向任务。毫不奇怪，考虑
appearing targets and to perform an orientation task requiring coordinating video movements 到上述发现，UGV图像的添加改善了识别。
with the map orientation. Not surprisingly, considering the aforementioned findings, the addition
of the UGV imagery improved identification.
Figure 6. Ben-Gurion University combined view display. 图6。本-古里安大学联合视图显示。
More interesting was the improvement in the orientation task in their first study because based 更有趣的是他们的第一项研究中定向任务的改进，因为基于先前的研究，Orphir等人。
on prior research, Orphir et al. predicted that UGVs would not be useful in aligning the videos 预测UGVs在将视频与地图位置对齐方面没有用。第二项研究验证了目标识别的组合视图
with the map locations. The second study verified the combined-views advantage for target 优势，但没有复制方向发现。这两个实验证明了将"窗外"UGVs的观点与UASs的"上帝之
identification but did not replicate the orientation finding. The two experiments demonstrated the 眼"观点相结合的好处。
benefits of combining the “out-of-window” UGVs’ view with the “god’s eye” view of UASs.
Further research showed that the efficacy of combining views depended on three factors: the mix 进一步的研究表明，组合视图的有效性取决于三个因素：图像的组合，参与者群体和显
of imagery, the participant population and, the simplicity of the displays (Oron-Gilad et al., in 示器的简单性(Oron-Gilad et al.，在准备中）。美国步兵士兵在英国《金融时报》的一项
preparation). U.S. infantry Soldiers in a field study at Ft. Benning made targeting and orientation 实地研究中. Benning通过具有UGV图像中心视图的显示器和具有场景附加视图的插入来做
decisions with a display with a center view of UGV imagery and inserts with additional views of 出定位和定向决策。插入有三种图像馈送条件：额外的UGV视图、UAS视图和具有单独
the scene. There were three imagery-feed conditions for inserts: additional UGV views, UAS 插入的UGV和UAS插入的组合视图。格式是一个单独的条件，涉及八种格式，其中包含
用于显示插入的各种方案，包括在视图之间切换。单独比较的八种格式不会影响性能；
views, and combined views with separate inserts of UGV and UAS inserts. Format was a
但是，图像可用性对定位和方向决策都有显着影响。因此，对于识别目标，显示具有多
separate condition and involved eight formats with various schemes for presenting the inserts,
个配置的配置
including toggling between views. The eight formats compared individually did not affect
performance; however, image availability had significant effects on both targeting and
orientation decisions. Thus for indentifying targets, display configurations that had multiple
15 15
UGV (“out of the window”) views improved performance. However, for inserts that displayed UGV（"窗外"）查看改进的性能。但是，对于显示多个UGV和UAS视图的插入，过多的
multiple UGV and UAS views, the plethora of choices actually degraded performance. Similar 选择实际上会降低性能。在定向决策方面，类似的结果是明显的，只有UGV的条件明显
results were evident for orientation decisions in that UGV-only conditions were significantly 优于所有具有UAS可用性的条件。总之，当监视处于操作场景的静止区域时（与Ophir-A
better than all conditions with UAS availability. To summarize, when the surveillance was in a rbelle等人相比。，2013年，更动态的场景），有更多的图像选择比最小的必要完成任
stationary area of an operations scenario (compared to Ophir-Arbelle et al., 2013, more dynamic 务降低性能。因此，美国士兵更喜欢并在所有UGV地面视图上做得最好，而当他们必须
同时将注意力集中在UAS和多个UGV图像上时，情况就不那么好。
scenarios), having more imagery choices than the minimal necessary to complete the tasks
degraded performance. Thus, U.S. Soldiers preferred and did best with all UGV ground views
and less well when they had to focus their attention on both UAS and multiple UGV imagery
simultaneously.
A subset of Oron-Gilad et al.’s study (in preparation) was rerun with BGU students who actually Oron-Gilad等人的子集。这项研究（准备中）是由BGU的学生重新进行的，他们实际上
had more combat arms experience (3+ years) than the Ft. Benning participants (1.9 years). The 比英国《金融时报》有更多的战斗武器经验（3年以上）。 Benning参与者（1.9年）。 B
BGU study only examined conditions where they could choose among additional ground and/or GU的研究只考察了他们可以在额外的地面和/或地面中选择的条件
UAS views with the UGV center view always available. In general, they did better than the Ft. UAS视图与UGV中心视图始终可用. 总的来说，他们比英国《金融时报》做得更好。本
Benning participants for these same conditions. BGU participants toggled to the UAS views 宁参与者这些相同的条件。 BGU参与者在大约75%的时间内切换到UAS视图，因此能够
approximately 75% of the time and thus were able to extract better information from the same 从相同的UAS视图中提取比经验较少的Ft更好的信息。本宁士兵。 Oron-Gilad等人。建
UAS views than the less experienced Ft. Benning Soldiers. Oron-Gilad et al. suggest that this is 议这是由于以色列国防军在兵役期间提供了更广泛的训练和空中地图的实际经验。因此
due to the IDF giving more extensive training and hands-on experience with aerial maps during ，影响性能的四个变量是经验，训练，图像和任务类型（与Ophir et al.的研究）。这些
系列实验与Ophir-Arbelle等人的差异。（2013）应注意的研究。 Ophir Arbelle等人。
their military service. Thus, the four variables that influenced performance were experience,
同时使用BGU学生，在拥有UGV和
training, imagery, and type of task (compared to Ophir et al.’s studies). Differences between
these series of experiments and the Ophir-Arbelle et al. (2013) studies should be noted. Ophir-
Arbelle et al., also using BGU students, showed a definite advantage for having both UGV and
UAS imagery compared to UAS imagery alone, whereas in the Oron-Gilad et al. BGU study, the UAS图像与单独的UAS图像相比，而在Oron-Gilad等人。 BGU研究，重点主要是如何呈
focus was predominantly on how to present additional imagery if the UGV imagery was 现额外的图像，如果UGV图像是可用的中心显示。
available on the center display.
The difference in utility of feeds seems to be due to the more dynamic quality of Ophir-Arbelle 饲料效用的差异似乎是由于Ophir-Arbelle等人的更动态质量。's（2013）研究。在这项
et al.’s (2013) study. In that study, the observers had to locate a moving target over a larger area 研究中，观察员必须在更大的区域上找到一个移动目标，而不是监视一个特定的位置。
rather than surveil a specific location. Also, most of Oron-Gilad et al.’s (in preparation) eight 此外，Oron-Gilad等人的大部分。八种格式（准备中）更复杂，可能有多达三个UGV窗口
formats were more complicated with the possibility of up to three UGV windows and a single 和一个具有各种切换方案和插入大小的UAS窗口。 Ophir-Arbelle研究使用了两个窗口，
UAS window with various toggling schemes and insert sizes. The Ophir-Arbelle study used two 两者都不需要切换（图6）。 Oron-Gilad（准备中）验证了在UGV和UAV视图之间切换（
即使具有更大显示表面的优势）对Ophir-Arbelle范式的目标识别具有有害影响。这些研
windows, neither of which required toggling (figure 6). Oron-Gilad (in preparation) verified that
究的含义是在使用这两种类型的图像时保持简单。此外，主显示器的设计应该足够大，
toggling between UGV and UAV views (even with the advantage of a larger display surface) had
以便操作员可以根据任务专注于其中一种或两种类型的图像。此外，拥有这两种类型的
a deleterious effect on target identification for the Ophir-Arbelle paradigm. The implication of 图像对于动态情况特别有用，其中一种图像类型或另一种图像类型由于UVs移动时的障碍
these studies is to keep it simple when using both types of imagery. Moreover, the design of the 物或其他因素而处于劣势（Ophir-Arbelle et al., 2012).
main displays should be large enough so that the operator can focus on either or both types of
imagery, depending on the mission. Also, having both types of imagery is particularly useful for
dynamic situations wherein one imagery type or another is at a disadvantage because of
obstructions or other factors as the UVs move (Ophir-Arbelle et al., 2012).
Another use of situation understanding imagery was explored by Evans (2012) and coresearchers 埃文斯（2012）和dcs公司的coresearchers探索了情境理解图像的另一个用途。他们研
at the DCS Corporation. They examined the utility of using overlays superimposed on the 究了使用叠加在
16 16
autonomous UGV imagery. The overlays (figure 7) were generated by the autonomous 自主UGV图像。复盖图（图7）由自主导航系统（ANS）生成，指示UGV未来路径的短期
navigation system (ANS) indicating a short-term prediction (a few seconds ahead) and a long- 预测（提前几秒钟）和长期预测（几分钟）。第一个实验是在ARL和DCS实验室进行的，
term prediction (a few minutes) of the UGV’s future path. The first experiment was conducted in 其中图像是由建模和模拟环境（MODSIM）生成的。 MODSIM是一种基于物理的仿真，
ARL and DCS laboratories wherein the imagery was generated by the Modeling and Simulation 用于实时模拟为自主UGV开发的ANS和传感器。在某些情况下，由激光护林员查找器生
Environment (MODSIM). MODSIM is a physics-based simulation that ARL used to simulate in 成的障碍物复盖被表示为图7中显示为"窗外"视图的红色和绿色区域。
real time the ANS and sensors developed for autonomous UGV. In some conditions, obstacle
overlays generated by the laser ranger finder were represented as red and green areas shown as
the “out-of-the-window” view in figure 7.
Figure 7. An example of the Warfighter machine interface (WMI) showing both the short-term 图7。显示短期（绿色）和长期（蓝色）操作员辅助工具的Warfighter机器接口（WMI
(green) and long-term (blue) operator aids. ）示例。
Combinations of obstacle and prediction overlays did not affect the operator’s ability to identify 障碍物和预测叠加的组合不影响操作员识别潜在目标、总任务时间、路线偏差或操作员
potential targets, total mission time, route deviations, or the number of times the operator 从ANS切换到远程操作的次数的能力。虽然17名参与者的监督控制相关分数不受叠加影
switched from the ANS to teleoperations. Whereas the 17 participants’ supervisory-control- 响，但操作员的主观评分和工作量估计都支持可视化的有效性（即更好的评分和更低的工
related scores were not affected by the overlays, the operator’s subjective ratings and workload 作量评分）。未能显示出显着的性能差异可能归因于各种原因，包括短暂的练习会话。 E
estimates both supported the visualization’s effectiveness (i.e., better ratings and lower workload vans（2012）认为，最可能的原因是为模拟选择的特定任务的便利性。具体而言，目标
很容易被发现（例如，笑脸），并且在大多数情况下，障碍同样明显。
scores). The failure to show significant performance differences could be attributed to a variety
of causes, including a brief practice session. Evans (2012) argues that the most likely causes
were the ease of the particular missions chosen for the simulations. Specifically, the targets were
easily spotted (e.g., smiley faces) and obstacles, for the most part, were equally obvious.
For his second study, Evans (under review) developed a more militarily relevant mission 对于他的第二项研究，埃文斯（正在审查中）开发了一个更具军事相关性的任务环境，
environment with realistic targets and more difficult obstacle decisions, such as whether to drive 具有现实的目标和更困难的障碍决策，例如是否在草地上驾驶。这次试验是在Lejeune营
over grass. It was conducted at Camp Lejeune, a Marine training base where the capstone 进行的,这是一个海洋训练基地,在城市地形地点的军事行动中进行了源的顶石试验。九
experiment for SOURCE was being held at the Military Operations in Urban Terrain site. Nine
17 17
Soldiers participated in the field experiment using the same MODSIM environment as in the 士兵们使用与实验室实验中相同的MODSIM环境参加了野外实验。实地研究在各种方面更
laboratory experiment. The field study was more realistic in a variety of ways: the travel planner 加现实：旅行规划师（TP）根据地形的不同，从几秒钟到几分钟的前方预览车辆路径，这
(TP) previewed the vehicular path from a few seconds to several minutes ahead depending on the 给了士兵类似于图7所示的自动驾驶车辆预测路径的视觉视图。还有一个障碍地图和一个
terrain, which gave the Soldiers visual views of the predicted path of the autonomous vehicle 警告导航变化的重新路由警报。 SA探测器向操作员询问了类似Ied的更现实的物体，并且
similar to those shown in figure 7. There was also an obstacle map and a rerouting alert that 障碍物图显示了与先前实验中更明显的桶状障碍物相比，操作员更难以辨别的草地区域。
与操作员减少对远程操作的依赖有关的相关措施显示了TP、TP+警报和TP+障碍图的显着
warned of navigation changes. The SA probes queried the operators concerning more realistic
影响。运营商在TP单独叠加并与其他辅助工具协同叠加时更多地依赖自治，这表明这些
objects resembling IEDs, and the obstacle map showed grass areas that were more difficult for
辅助工具在提高适当信任方面的优势（Lee和See，2004）。也就是说，当他们有视觉增
the operator to discern compared to the more obvious barrel obstacles in the previous 强预览预测的自主路线时，他们更频繁地依赖于自主性。
experiment. The dependent measures relating to the operator reducing his or her reliance on
teleoperations showed significant effects of the TP alone, TP + alert, and TP + obstacle map. The
operators’ greater reliance on autonomy when they had TP overlays both alone and in concert
with other aids shows the advantage of these aids in improving appropriate trust (Lee and See,
2004). That is, they relied on autonomy more often when they had visual augmentation
previewing the projected autonomous route.

• The requirement for video display size varied depending on the mission, terrain, clutter, • 视频显示尺寸的要求因任务、地形、杂波和任务而异。精细颗粒分析需要7.5英寸
and task. Video displays of 7.5 in were required for fine-grain analysis especially in urban 的视频显示,特别是在城市地区. 然而,3.5英寸的视频显示器在要求较低的环境中就足
areas. However, 3.5-in video displays were sufficient in less demanding environments. 够了.
• In two experiments, Ben-Gurion University using Israeli participants with at least 3-years • 在两项实验中，本-古里安大学使用具有至少3年军事经验的以色列参与者发现，U
military experience found that a combination of UAS and UGV imagery improved their AS和UGV图像的组合改善了他们的监视性能。
surveillance performance.
• U.S. Soldiers at Ft. Benning preferred UGV imagery to UAS imagery and actually showed • 英国《金融时报》的美国士兵。 Benning更喜欢UGV图像而不是UAS图像，并且实
degraded performance with too many UGV and UAS imagery inserts, indicating that too 际上显示出过多的UGV和UAS图像插入的性能下降，这表明过多的图像是有害的。
much imagery was detrimental.
• Israeli participants with more extensive experience with aerial views preferred UAS • 具有更多鸟瞰经验的以色列参与者更喜欢UAS图像。在UAV和UGV视图之间来回切
imagery. Toggling back and forth between UAV and UGV views also lead to poorer target 换也会导致较差的目标识别。
identification.
• The general conclusion was that two heterogeneous sources of imagery were better than • 总的结论是，两个不同的图像来源比一个更好，特别是对于动态监视。偏好取
one, especially for dynamic surveillance. Preferences depended on previous experience 决于以前使用UV图像的经验；对于最简单的显示配置，可以获得最佳性能。
with UV imagery; the best performance was obtained for the simplest display
configurations.
• Augmented interfaces that showed the predicted path of the autonomous UV, alerted the • 增强的界面显示自主UV的预测路径，提醒操作员导航变化，并指出障碍物的位置
operator to navigational changes, and indicated locations of obstacles improved the ，提高了操作员的信任度。增强的辅助设备显着减少了操作员从自动切换到传送
operator’s trust. The augmented aids significantly reduced the number times and the 模式的实例的次数和持续时间。
durations of the instances that the operator switched from the autonomous to the teleop
mode.
18 18
7. Conclusions and Future Research 7. 结论和未来研究
Multiple sources indicate a future in which autonomy will become an important part of our lives 多种来源表明，未来自治将成为我们生活的重要组成部分，也是现代战争的重要组成部
as well as an important component of modern warfare (Barnes and Chen, 2012; Barnes and 分（Barnes和Chen，2012;Barnes和Evans，2010;Chen et al。，2011b;Chen和Barnes，
Evans, 2010; Chen et al., 2011b; Chen and Barnes, in press; Weiss, 2011). However, the human in press;Weiss，2011）。然而，人类的角色不会减少，但会变得更像高管，设定更长期
role will not diminish but will become more executive-like, setting longer-term goals and 的目标，并干预安全，自主故障或由于不断变化的目标（Barnes和Evans，2010;Chen et
intervening for safety, autonomy malfunctions, or because of changing objectives (Barnes and al。, 2007). 在可预见的未来，人类的元知识和道德责任将要求自治系统在生命处于危险
中时服从于人类的决策权威（Chen和Barnes，in press）。
Evans, 2010; Chen et al., 2007). For the foreseeable future, the human’s meta-knowledge and
ethical responsibilities will require that autonomous systems defer to human decision authority
when lives are at risk (Chen and Barnes, in press).
This report covered findings in a variety of experimental paradigms pertaining to the central 这份报告涵盖了与自治核心问题有关的各种实验范式的发现：有多少干预是必要的，在
issue of autonomy: how much intervention is necessary and under what conditions should 什么条件下人类应该信任自治而不是干预。虽然研究结果很复杂，但Jentsch及其同事(20
humans trust the autonomy and not intervene. Although the findings were complex, Jentsch and 12)报告说，一般来说，让人类参与循环对于改善SA是积极的，或者因为人类比自治实体
colleagues (2012) reported that, in general, having the human in the loop was positive either for （通过代理感知）更好地执行特定功然而，在某些情况下，人为干预会适得其反，要么
improved SA or because the humans performed a specific function better than the autonomous 是因为时间限制，要么是因为工作量要求对人的干预能力产生了负面影响。不幸的是，
也有证据表明，在危险情况下，人类因为自满或忽视而过度信任自动化，导致灾难性事故
entity (perception by proxy). However, there were conditions in which human intervention was
（Parasuraman和Riley，1997）。自适应控制被认为是一种可能的补救措施，因为它允
counterproductive, either because of time constraints or because the workload requirements
许人类在中等工作量下处于循环中，并且当多任务困难阻碍人类表现时，它允许自动控制
negatively affected the human’s ability to intervene. Unfortunately, there is also evidence that in (Barnes et al., 2006).
dangerous situations, humans over-trust automation because of complacency or neglect, resulting
in catastrophic accidents (Parasuraman and Riley, 1997). Adaptive control was suggested as a
possible remedy because it allowed the human to be in the loop under moderate workload, and it
allowed automated control when multitasking difficulty impeded human performance (Barnes
et al., 2006).
The RoboLeader research suggests that intelligent agents are possible solutions to the emerging RoboLeader的研究表明，智能代理是让单个操作员监督多个半自治系统这一新兴问题的
problem of having a single operator supervise multiple semi-autonomous systems. The 可能解决方案。 RoboLeader代理是一个中间主管，控制能力较弱的半自治系统，并反过
RoboLeader agent is an intermediate supervisor that controls less capable semi-autonomous 来在人类操作员的监督下。当出现问题时，代理会通知操作员该问题，并在操作员的许
systems and is in turn under the supervision of the human operator. When something is amiss, 可下执行解决方案。这允许操作员在高工作量任务期间进行多任务处理，并且仍然保持
the agent informs the operator of the issue and executes a solution with the operator’s 决策权限;然而，决策权限可以在指定条件下委派给代理人，例如时间关键的紧急情况（C
hen和Barnes，in press）。实验结果表明，使用代理作为中介是控制多个机器人的有效
permission. This allows the operator to multitask during high-workload missions and still
方法，特别是在困难的任务情况下（甚至在代理不完全可靠的条件下）（Chen和Barnes
maintain decision authority; however, decision authority can be delegated to agents under
，2012a;2012b）。特别令人感兴趣的是发现操作员的个体差异是人机性能的重要决定
specified conditions, such as a time-critical emergency (Chen and Barnes, in press). The 因素（Chen，2011）。
experimental results suggest that using an agent as an intermediary is an efficient way to control
multiple robots, especially under difficult tasking situations (and even under conditions in which
the agent is less than perfectly reliable) (Chen and Barnes, 2012a; 2012b). Of particular interest
is the finding that operators’ individual differences are an important determinate of human-agent
performance (Chen, 2011).
19 19
The report also examined advanced interfaces for small robots for both autonomous and 该报告还研究了在Ft现场实验期间用于自主和远程操作条件的小型机器人的高级接口。本
teleoperated conditions during field experiments at Ft. Benning. The purpose of the field studies 宁。实地研究的目的是调查释放操作员双手、增加SA和改善任务安全的接口。自主SUG
was to investigate interfaces that freed the operator’s hands, increased SA, and improved mission Vs能够更快地从a点移动到B点，减少错误，同时围绕中间目标导航并避开障碍物。部分
safety. Autonomous SUGVs were able to move from point A to point B more rapidly with fewer 自治被证明是令人沮丧的，并不像完全自治那样有利。其他研究表明，语音控制显示出
errors while navigating around intermediate objectives and avoiding obstacles. Partial autonomy 希望，但仍然需要控制增强，以便有效地进行连续的课程校正，而临场感研究表明，在远
程观看期间使用3-D音频线索进行目标检测的操作员具有优势。同样，在Ft进行现场实验
proved frustrating and not nearly advantageous as full autonomy. Other research indicates that
。 Leonard Wood展示了将SUGV立体视觉与触觉臂操纵器相结合的实用性，用于IED任务
voice control shows promise but still requires control augmentation in order to make continuous
。
course corrections efficiently, whereas telepresence research indicates an advantage for operators
using 3-D audio cues for target detection during remote viewing. Similarly, field experiments at
Ft. Leonard Wood demonstrated the utility of combining SUGV stereovision with a haptic arm
manipulator for IED missions.
Finally, our Israeli colleagues working with Ft. Benning researchers investigated the Soldier’s 最后，我们的以色列同事与英国《金融时报》合作。本宁的研究人员调查了这名士兵使
ability to use multiple UVs to conduct surveillance and reconnaissance missions. The findings 用多个Uv进行监视和侦察任务的能力。调查结果很复杂，表明任务动态，军事经验和显
were complex, indicating that mission dynamics, military experience, and display characteristics 示特征都有助于成功的监视任务。最普遍的发现是，只要显示配置简单，UAS和UGV图
all contributed to successful surveillance missions. The most general findings are that UAS and 像就可以协同组合，并且图像与操作员的军事经验一致。这项研究很重要，因为自治系
UGV imagery can be combined synergistically as long as the display configuration is simple, 统最有前途的用途之一是持续监控，将多个图像源广播到安全位置。此外，我们的内部
研究表明，增强的状态信息，例如自动驾驶车辆的预测路径，增加了士兵的信任，因为
and the imagery is congruent with the operators’ military experience. This research was
当系统有效运行时，他或她不太可能干扰自治。
important because one of the most promising uses of autonomous systems is persistent
surveillance with multiple sources of imagery being broadcast to a safe location. Also, our in-
house research indicates that augmented status information, such as the projected path of the
autonomous vehicle, increases the Soldier’s trust in the sense that he or she is less likely to
interfere with autonomy when the system is performing effectively.
There are two major thrusts for future research. UV interfaces are becoming more naturalistic, 未来的研究主要有两个方面。 UV界面变得更加自然，从而模仿士兵与其他士兵的互动。
thus emulating the Soldier’s interaction with other Soldiers. The future Soldier will be able to 未来的士兵将能够更自然地使用来自远程UVs的3-D音频和视觉临场感信息与自治系统一
work more naturally with autonomous systems using 3-D audio and visual telepresence 起工作。此外，未来的士兵将能够使用手势，触觉和语音界面重定向Uv。我们的研究目
information from remote UVs. Also, future Soldiers will be able to redirect UVs using gesture, 标将是使士兵能够毫不费力地结合这些技术，而不会增加已经携带的重量或干扰其他任务
haptic, and voice interfaces. Our research objectives will be to enable Soldiers to effortlessly 要求。自然界面的承诺是在激烈的战斗中对多个系统进行免提和眼睛向前控制。
combine these technologies without adding to the weight already carried or interfering with other
mission requirements. The promise of naturalistic interfaces is a hands-free and eyes-forward
control of multiple systems during the heat of combat.
The second thrust involves intelligent agents. Most of the intelligent systems involved in 第二个推力涉及智能代理。大部分涉及的智能系统
SOURCE incorporated sensors and algorithms to perform behaviors autonomously, such as SOURCE集成了传感器和算法，以自主执行行为，例如进行限制任务所必需的更改路线或
rerouting or pedestrian avoidance necessary to conduct circumscribed missions. True intelligence 避免行人。真正的智能更广泛，涉及适应新环境，推断意图以及与其他智能实体沟通（C
is broader and involves adapting to new environments, inferring intent, and communicating with hen和Barnes，in press）。这将需要至少具有模仿人类对应物的有限智力的代理人。我
other intelligent entities (Chen and Barnes, in press). This will require agents that have at least 们的研究工作（与其他ARL赞助和国防部研究团队合作）将专注于人-代理人互动的两个
limited intelligence that mimics their human counterparts. Our research efforts (working with 方面：代理人透明度和双向沟通。第一个目标将是发展透明度原则和原型显示，以加强
other ARL-sponsored and Department of Defense research teams) will focus on two aspects of
human-agent interaction: agent transparency and bidirectional communication. The first
objective will be to develop transparency principles and prototype displays that enhance the
20 20
operator’s ability to understand the agent’s current state, predict future agent states, and infer the 操作员了解代理当前状态、预测未来代理状态以及推断代理行为意图的能力。双向性意
intent of the agent’s behavior. Bidirectionality implies that the agent can communicate with 味着代理人可以与人类就人类的意图，军事环境的可能变化和澄清进行沟通。双向性将
humans concerning the human’s intent, possible changes in the military environment, and 需要图形，注释图像，手势和人与代理人之间的口头交流。具有更大自主权的智能代理
clarifications. Bidirectionality will entail graphical, annotated imagery, gestures, and verbal 和无人系统将改变军事格局。然而，它不会使士兵的任务更容易;它只会改变他或她将面
communication between humans and agents. Intelligent agents and unmanned systems with 临的任务类型和问题（Chen和Barnes，in press）。
greater autonomy will change the military landscape. However, it will not make the Soldier’s
task easier; it will only change the type of tasks and problems he or she will face (Chen and
Barnes, in press).
21 21
8. References 8. References
Barnes, M. J.; Chen, J. Y. C. Intelligent Agents as Supervisory Assets for Multiple Uninhabited Barnes,M.J.;Chen,J.Y.C.智能代理作为多个无人居住的监管资产
Systems: RoboLeader. In Supervisory Control of Multiple Uninhabited Systems- 系统:RoboLeader. 在多个无人居住系统的监督控制
Methodologies and Enabling Human-Robot Interface Technologies; RTO-TR-HFM-170; Methodology and Enabling Human-Robot Interface Technologies;rto-TR-HFM-170;
van Breda, L., Ed.; NATO Science and Technology Organization: Neuilly sur Seine, France, van Breda,L.,Ed.;北约科学技术组织：法国塞纳河畔讷伊，2012;pp16-1-16-14。
2012; pp 16-1–16-14.
Barnes, M. J.; Chen, J. Y. C.; Jentsch, F.; Redden, E. S. Designing Effective Soldier-Robot Barnes,M.J.;Chen,J.Y.C.;Jentsch,F.;Redden,E.S.设计有效的士兵机器人
Teams in Complex Environments: Training, Interfaces, and Individual Differences. In 复杂环境中的团队：培训、界面和个体差异。在
Engineering Psychology and Cognitive Ergonomics, Lecture Notes in Computer Science; 工程心理学与认知人机工程学，计算机科学讲义 ;
Harris, D., Ed.; Springer-Verlag: Berlin, 2011; Vol. 6781, pp 484–493. 哈里斯，D。，Ed。;Springer-Verlag：柏林，2011;Vol。 6781，第484-493页。
Barnes, M. J.; Evans, A. W., III. Soldier-Robot Teams in Future Battlefields: An Overview. In 巴恩斯，M.J.;埃文斯，A.W.，III。未来战场的士兵-机器人团队：概述。在
Human-Robot Interactions in Future Military Operation; Barnes, M. J., Jentsch, F., Eds.; 未来军事行动中的人-机器人交互;Barnes，M.J.，Jentsch，F.，Eds。;
Ashgate: Farnham, Surrey, UK, 2010; pp 9–29. Ashgate：Farnham，Surrey，UK，2010;pp9-29。
Barnes, M. J.; Grossman, J. The Intelligent Assistant Concept for Electronic Warfare Systems; Barnes,M.J.;Grossman,J.电子战系统的智能助理概念 ;
Technical Report NWC TP 5585; Naval Warfare Center: China Lake, CA, 1985. 技术报告NWC TP5585;海战中心:中国湖,CA,1985.
Barnes, M. J.; Parasuraman, R.; Cosenzo, K. Adaptive Automation for Robotic Military Systems. Barnes，M.J.;Parasuraman，R.;Cosenzo，K.机器人军事系统的自适应自动化。
Human Factors Issues for Uninhabited Military Vehicle; NATO Publication No. RTOTR- 无人居住的军用车辆的人为因素问题；北约出版物第1号。 RTOTR
HFM-O78; NATO Science and Technology Organization: Neuilly sur Seine, France, 2006; HFM-O78;北约科学技术组织：法国塞纳河畔讷伊，2006;pp420-440。
pp 420–440.
Beck, H.; Dzindolet, M.; Pierce, L. Automation Usage Decisions: Controlling Intent and Beck，H.;Dzindolet，M.;Pierce，L.自动化使用决策：控制意图和
Appraisal Errors in a Target Detection Task. Human Factors 2007, 49, 429–437. 目标检测任务中的评估错误。人的因素2007,49,429-437。
Chen, J. Y. C. Individual Differences in Human-Robot Interaction in a Military Multitasking Chen,J.Y.C.军事多任务中人机交互的个体差异
Environment. (Special Issue on Improving Human-Robot Interaction in Complex 环境。 (综合体改善人机互动特刊)
Operational Environments: Translating Theory into Practice.) Journal of Cognitive 操作环境：将理论转化为实践。）认知杂志
Engineering and Decision Making 2011, 5 (1), 83–105. 工程与决策2011，5（1），83-105。
Chen, J. Y. C. Effects of Operators’ Spatial Ability on Their Performance of Target Detection Chen,J.Y.C.操作员空间能力对目标检测性能的影响
and Robotics Tasks. In Designing Soldier Systems: Current Issues in Human Factors (series 和机器人任务。 In Designing Soldier Systems:Current Issues in Human Factors(Serie
of Human Factors in Defence; Savage-Knepshield, P., Martin, J., Lockett, J., Allender, L., s Of Human Factors in Defense;Savage-Knepshield,P.,Martin,J.,Lockett,J.,Allender,L.,
Eds.; Ashgate: Farnham, Surrey, UK, 2012; pp 53–67. Eds.;Ashgate：Farnham，Surrey，UK，2012;pp53-67。
Chen, J. Y. C.; Barnes, M. J. Supervisory Control of Multiple Robots: Effects of Imperfect Chen,J.Y.C.;Barnes,M.J.多个机器人的监督控制：不完美的影响
Automation and Individual Differences. Human Factor 2012a, 54 (2), 157–174. 自动化和个体差异。人因子2012a，54（2），157-174。
Chen, J. Y. C.; Barnes, M. J. Supervisory Control of Multiple Robots in Dynamic Tasking Chen,J.Y.C.;Barnes,M.J.动态任务中多个机器人的监督控制
Environments. Ergonomics 2012b, 55 (9), 1043–1058. 环境。人体工程学2012b，55（9），1043-1058。
22 22
Chen, J. Y. C.; Barnes, M. J.; Harper-Sciarini, M. Supervisory Control of Multiple Robots: Chen,J.Y.C.;Barnes,M.J.;Harper-Sciarini,M.多个机器人的监督控制:
Human Performance Issues and User Interface Design. IEEE Transactions on System, Man, 人的性能问题和用户界面设计。 IEEE Transactions on System,Man,And Cybernetics
and Cybernetics—Part C: Applications and Reviews 2011a, 41 (4), 435–454. -Part C:Applications and Reviews2011a,41(4),435-454.
Chen, J. Y. C.; Barnes, M. J. Human-Agent Teaming for Multi-Robot Control: A Review of Chen,J.Y.C.;Barnes,M.J.Human-Agent Teaming for Multi-Robot Control:A Review of
Human Factors Issues. In IEEE Transactions on Human-Machine Systems (in press). 人为因素问题。在IEEE Transactions on Human-Machine Systems(in press)中。
Chen, J. Y. C.; Barnes, M. J.; Quinn, S. A.; Plew, W. Effectiveness of RoboLeader for Dynamic Chen,J.Y.C.;Barnes,M.J.;Quinn,S.A.;Plew,W.Roboleader对动态的有效性
Re-Tasking in an Urban Environment. Proceedings Human Factors and Ergonomics Society 在城市环境中重新任务。人类因素和人体工程学会第五十五届年会，拉斯维加斯，
55th Annual Meeting, Las Vegas, NV, 19–23 September 2011b; pp 1501–1505. NV，2011年9月19-23日b；第1501-1505页。
Chen, J. Y. C.; Haas, E. C.; Barnes, M. J. Human Performance Issues and User Interface Design Chen,J.Y.C.;Haas,E.C.;Barnes,M.J.远程操作机器人的人类性能问题和用户界面设计。 IEEE T
for Teleoperated Robots. IEEE Transactions on Systems, Man, and Cybernetics—Part C: ransactions on Systems,Man,And Cybernetics-Part C:Applications and Reviews2007,37
Applications and Reviews 2007, 37, 1231–1245. ,1231-1245.
Edmondson R.; Light K.; Bodenhamer A.; Bosscher P.; Wilkinson L. Enhanced Operator Edmondson R.;Light K.;Bodenhamer A.;Bosscher P.;Wilkinson L.增强型操作员
Perception Through 3D Vision and Haptic Feedback. Proceedings of SPIE, Unmanned 通过3D视觉和触觉反馈感知。无人驾驶的SPIE会议记录
Systems Technology XIV, Baltimore, MD, 2012; Vol. 8373, 25–27. Systems Technology XIV，Baltimore，MD，2012;Vol。 8373, 25–27.
Elliott, L. R.; Jensen, C.; Redden, E. S.; Pettit, E. Robotic Telepresence: Perception, Elliott，L.R.;Jensen，C.;Redden，E.S.;Pettit，E.Robotic Telepresence：Perception,
Performance and User Experience; ARL-TR-5928; U.S. Army Research Laboratory: 性能和用户体验;ARL-TR-5928;美国陆军研究实验室:
Aberdeen Proving Ground, MD, 2012. 阿伯丁试验场,MD,2012.
Evans, A. W. Investigating the Usefulness of Operator Aids for Autonomous Unmanned Ground Evans,A.W.调查自动无人地面操作员辅助设备的有用性
Vehicles Performing Reconnaissance Tasks; ARL-TR-6651; U.S. Army Research 执行侦察任务的车辆;ARL-TR-6651;美国陆军研究
Laboratory: Aberdeen Proving Ground, MD, 2012. 实验室:阿伯丁试验场,MD,2012.
Evans, A.W. Investigating the Usefulness of Soldier Aids for Autonomous Unmanned Vehicles, 埃文斯，A.W.调查士兵辅助自动无人驾驶车辆的有用性,
Part 2 (under review). Part 2 (under review).
Fincannon, T.; Evans, A. W., III; Jentsch, F.; Keebler, J. Dimensions of Spatial Ability and Their Fincannon，T.;Evans，A.W.，III;Jentsch，F.;Keebler，J.空间能力的维度及其
Influence on Performance With Unmanned Systems. In Human Factors in Defense: Human 对无人系统性能的影响。在防御中的人为因素：战斗识别中的人为因素;Andrews，
Factors in Combat Identification; Andrews, D. H., Herz, R. P., Wolf, M. B., Eds.; Ashgate: D.H.，Herz，R.P.，Wolf，M.B.，Eds。;阿什盖特:
Aldershot, UK, 2010; pp 67–81. Aldershot, UK, 2010; pp 67–81.
Fincannon, T.; Jentsch, F.; Sellers, B.; Keebler, J. R. Beyond “Spatial Ability”: Examining the Fincannon，T.;Jentsch，F.;Sellers，B.;Keebler，J.R.超越"空间能力"：检查
Impact of Multiple Individual Differences in a Perception-by-Proxy Framework. 代理感知框架中多个个体差异的影响。
Proceedings of the 7th ACM/IEEE International Conference on Human-Robot Interaction, 第七届ACM/IEEE人机交互国际会议论文集 ,
Boston, MA, March 2012. 波士顿，马，2012年3月。
Fincannon, T.; Keebler, J. R; Jentsch, F.; Phillips, E.; Evans, A. W. Team Size, Team Role, Fincannon，T.;Keebler，J.R;Jentsch，F.;Phillips，E.;Evans，A.W.团队规模，团队角色,
Communication Modality, and Team Coordination in the Distributed Operation of Multiple 多个分布式操作中的沟通方式和团队协调
Heterogeneous Unmanned Vehicles. Journal of Cognitive Engineering and Decision Making 异构无人车。认知工程与决策杂志2011，5（1），106-131。
2011, 5 (1), 106–131.
23 23
Goodrich, M. On Maximizing Fanout: Towards Controlling Multiple Robot Teams. In Human- Goodrich，M.关于最大化扇出：控制多个机器人团队。在人类
Robot Interactions in Future Military Operations; Barnes, M. J., Jentsch, F., Eds.; Ashgate, 未来军事行动中的机器人交互;Barnes，M.J.，Jentsch，F.，Eds。;阿什盖特,
Farnham, Surrey, UK, 2010; pp 368–393. Farnham，英国萨里，2010;pp368-393。
Goodrich, M.; Blatter, K. Temporal Latency in HRI. Final Report- Computer Science Goodrich，M.;Blatter，K.Hri中的时间延迟。最终报告计算机科学
Department; Brigham Young University: Provo, UT, 2012. 系;杨百翰大学：PROVO，UT，2012。
Jansen, C.; van Breda, L.; Elliott, L. Remote Auditory Target Detection Using an Unmanned Jansen，C.;van Breda，L.;Elliott，L.使用无人驾驶的远程听觉目标检测
Vehicle – Comparison Between a Telepresence Headtracking 3D Audio Setup and a 车辆-远程呈现Headtracking3D音频设置与带有定向麦克风的操纵杆控制系统之
Joystick-Controlled System With a Directional Microphone. In Supervisory Control of 间的比较。在多个无人居住系统的监督控制.方法和实现人-机器人接口
Multiple Uninhabited Systems-Methodologies and Enabling Human-Robot Interface
Technologies; van Breda, L., Ed.; RTO-TR-HFM-170; NATO Science and Technology Technologies;van Breda，L.，Ed。;RTO-TR-HFM-170;北约科技
Organization: Neuilly sur Seine, France, 2012; pp 11-1–11-14. 组织：法国塞纳河畔讷伊，2012年；第11-1-11-14页。
Jentsch, F.; Fincannon, T. Multiple Robot-Multiple Operator Control and Teamwork. Jentsch，F.;Fincannon，T.多个机器人-多个操作员控制和团队合作。
Proceedings of the Information Technology Conference of the American Institute for 美国航空航天学会信息技术会议记录（"Infotech@AIAA"），加利福尼亚州奥
Aeronautics and Astronautics (“Infotech @ AIAA”), Orange County, CA, June 2012. 兰治县，2012年6月。
Jentsch, F.; Fincannon, T.; Ososky, S.; Keebler, J.; Phillips, E. Effects of Robot Teaming and Jentsch，F.;Fincannon，T.;Ososky，S.;Keebler，J.;Phillips，E.机器人组队和
Reliability on Operator Supervisory Control; University of Central Florida: Orlando, FL, 操作员监督控制的可靠性;中佛罗里达大学:奥兰多,FL,2010.
2010.
Jentsch, F.; Fincannon, T.; Sellers, B.; Keebler, J.; Osoky, S.; Phillips, E.; Schuster, D. Safe Jentsch，F.;Fincannon，T.;Sellers，B.;Keebler，J.;Osoky，S.;Phillips，E.;Schuster，D.Safe
Operations of Autonomy, Teaming and Workload Operator Trust and Performance; UCF 自主、团队和工作负载操作员信任和性能的操作；UCF
Technical Report TO 132- 12/11; University of Central Florida: Orlando, FL, 2011. 技术报告13212/11;中佛罗里达大学：奥兰多，佛罗里达州，2011。
Lee, J. D.; See, K. A. Trust in Technology: Designing for Appropriate Reliance. Human Factors Lee，J.D.;参见，K.A.Trust in Technology：Designing For Appropriate Reliance。人的因
2004, 46, 50–80. 素2004，46，50-80。
Miller, C.; Parasuraman, R. Designing for Flexible Interaction Between Humans and 米勒，C.;Parasuraman，R.设计人类和人类之间的灵活互动
Automation: Delegation Interfaces for Supervisory Control. Human Factors 2007, 49, 自动化：用于监督控制的委派接口。人的因素2007，49，57-75。
57–75.
Miller, G. The Magical Number 7 Plus or Minus Two: Some Limits on our Capacity for 神奇的数字7加或减两：对我们的能力的一些限制
Processing Information. Psychological Review 1956, 63, 81–97. 处理信息。心理评论1956，63，81-97。
Mitchell, D.; Chen, J. Y. C. Impacting System Design With Human Performance Modeling and Mitchell,D.;Chen,J.Y.C.通过人的性能建模和
Experiments: Another Success Story. Proceedings of the Human Factors and Ergonomics 实验：另一个成功的故事。人类因素与人体工程学论文集
Society 50th Annual Meeting; HFES: San Monica, CA, 2006. 学会第50届年会;HFES:圣莫尼卡,CA,2006.
Ophir-Arbelle, R.; Oron-Gilad, T.; Borowsky, A.; Parmet, Y. Is More Information Better? How Ophir-Arbelle，R.;Oron-Gilad，T.;Borowsky，A.;Parmet，Y.更多信息更好吗？如何
Dismounted Soldiers Use Video Feed From Unmanned Vehicles: Allocation and 下马的士兵使用无人驾驶车辆的视频饲料：分配和
Information Extraction Considerations. Journal of Cognitive Engineering and Decision 信息提取注意事项。认知工程与决策杂志
Making 2013, 7 (1), 26–48. Making 2013, 7 (1), 26–48.
24 24
Oron-Gilad, T. When More Information Is Better; Multiple Video Feed Display Layout Oron-Gilad，T.当更多信息更好;多个视频馈送显示布局
Considerations, in preparation. 考虑，在准备。
Oron-Gilad, T.; Redden, E. S.; Minkov, Y. Robotic Displays for Dismounted Warfighters: A Oron-Gilad，T.;Redden，E.S.;Minkov，Y.下马战斗机的机器人显示器：A
Field Study. Journal of Cognitive Engineering and Decision Making (Special Issue on 实地考察。认知工程与决策杂志（关于
Improving Human-Robot Interaction in Complex Operational Environments: Translating 改善复杂操作环境中的人机交互：翻译
Theory into Practice, Part I) 2011, 5, 29–54. 理论转化为实践，第一部分）2011，5，29-54。
Oron-Gilad, T.; Redden, E. S.; Ophir-Arbelle, R. Are More Views Better? How Dismounted Oron-Gilad，T.;Redden，E.S.;Ophir-Arbelle，R.更多的观点更好吗？如何下马
Soldiers Use Video Feed From Unmanned Vehicles: Multiple Sensors and Display Layout 士兵使用无人驾驶车辆的视频馈送：多个传感器和显示布局
Considerations, in preparation. 考虑，在准备。
Pettitt, R. A.; Redden, E. S.; Carstens, C. B.; Hooper, D. Scalability of Robotic Controllers: Pettitt，R.A.;Redden，E.S.;Carstens，C.B.;Hooper，D.Robotic Controllers的可扩展性:
Effects of Progressive Autonomy on Intelligence, Surveillance, and Reconnaissance Robotic 渐进自治对情报、监视和侦察机器人的影响
Tasks; ARL-TR-6182; U.S. Army Research Laboratory: Aberdeen Proving Ground, MD, 任务;ARL-TR-6182;美国陆军研究实验室：阿伯丁试验场，MD，2012。
2012.
Pettitt, R.; Redden, E.; Pacis, E.; Carstens, C. Scalability of Robotic Controllers: Effects of Pettitt，R.;Redden，E.;Pacis，E.;Carstens，C.机器人控制器的可扩展性：
Progressive Levels of Autonomy on Robotic Reconnaissance Tasks; ARL-TR-5258; U.S. 机器人侦察任务的渐进自治水平;ARL-TR-5258;美国
Army Research Laboratory: Aberdeen Proving Ground, MD, 2010. 陆军研究实验室：阿伯丁试验场，MD，2010。
Phillips, E.; Fincannon, T.; Leotaud, P.; Jentsch, J. Effects of Robot Teaming and Operator Phillips，E.;Fincannon，T.;Leotaud，P.;Jentsch，J.机器人组队和操作员的影响
Memory on the Rate of Human Interaction With Multiple Unmanned Vehicles. Proceedings 人与多个无人驾驶车辆的交互速率的记忆。第27届陆军科学会议记录，奥兰多，佛
of the 27th Army Science Conference, Orlando, FL, 2010. 罗里达州，2010年。
Parasuraman, R.; Riley, V. Humans and Automation: Use, Misuse, Disuse, Abuse. Human Parasuraman，R.;Riley，V.人类和自动化：使用，误用，废弃，滥用。人类
Factors 1997 39, 230–253. Factors 1997 39, 230–253.
Parasuraman, R.; Barnes, M.; Cosenzo, K. Adaptive Automation for Human-Robot Teaming in Parasuraman，R.;Barnes，M.;Cosenzo，K.人机协作的自适应自动化
Future Command and Control Systems. The Int. C2 Journal 2007, 1, 43–68. 未来的指挥和控制系统。的Int。 C2期刊2007，1，43-68。
Redden, E.; Elliott, L.; Barnes, M. Robots: The New Team Members. In The Psychology of Redden，E.;Elliott，L.;Barnes，M.Robots：新团队成员。在心理学
Workplace Technology; Coovert, M., Thompson, L., Eds.; Society of Industrial 工作场所技术;Coovert，M.，Thompson，L.，Eds。;工业协会
Organizational Psychology Frontiers Series. Taylor and Francis: London, 2013; pp 185–208. 组织心理学前沿系列。泰勒和弗朗西斯：伦敦，2013;pp185-208。
Russell, S. J.; Norvig, P. Artificial Intelligence: A Modern Approach, 3rd ed.; Upper Prentice Russell，S.J.;Norvig，P.人工智能：现代方法，3rd ed。;上普伦蒂斯
Hall: Saddle River, NJ, 2009. 大厅:马鞍河,新泽西州,2009.
Sellers, B.; Fincannon, T.; Jentsch, F. The Effects of Autonomy and Cognitive Abilities on Sellers，B.;Fincannon，T.;Jentsch，F.自主和认知能力对
Workload and Supervisory Control of Unmanned Systems. Proc. Human Factors and 无人系统的工作量和监督控制。 Proc. 人为因素及
Ergonomics Society Annual Meeting, 2012; pp 1039–1043. 人体工程学会年会，2012年；第1039-1043页。
Weiss, L. G. Autonomous Robots in the Fog of War. IEEE Spectrum, August 2011. Weiss，L.G.战争迷雾中的自主机器人。 IEEE频谱，2011年8月。
25 25
Wickens, C. D.; Levinthal, B.; Rice, S. Imperfect Reliability in Unmanned Air Vehicle Wickens，C.D.;Levinthal，B.;Rice，S.无人驾驶飞行器中的不完美可靠性
Supervision and Control. In Human-Robot Interactions in Future Military Operations, 监督和控制。在未来军事行动中的人机交互 ,
Barnes, M. J., Jentsch, F., Eds.; Ashgate: Farnham, Surrey, UK, 2010; pp 211–228. Barnes，M.J.，Jentsch，F.，Eds。;Ashgate：Farnham，Surrey，UK，2010;pp211-228。
Zapotoczny, W. S., Jr. Shadow Warriors: Pennsylvania ARNG Tactical UAS Training Program. Zapotoczny，W.S.，Jr.影子战士：宾夕法尼亚州ARNG战术UAS训练计划。
Army Aviation 2007, 36–41; http://www.wzaponline.com/Shadow.pdf (accessed December 陆军航空2007，36-41;http://www.wzaponline.com/Shadow.pdf （2013年12月访
2013). 问）。
26 26
List of Symbols, Abbreviations, and Acronyms 符号、缩写和首字母缩写词列表
2-D two dimensional 2-D 二维
3-D three dimensional 3-D 三维立体
AA adaptive automation AA 自适应自动化
ANS autonomous navigation system ANS 自主导航系统
ARL U.S. Army Research Laboratory ARL 美国陆军研究实验室
AU autonomous AU autonomous
BGU Ben-Gurion University BGU Ben-Gurion University
FAP false alarm prone FAP 容易出现假警报
HHD handheld display HHD 手持式显示器
HMD helmet-mounted display HMD helmet-mounted display
HRED Human Research and Engineering Directorate HRED 人类研究及工程局
HRI human-robot interaction HRI human-robot interaction
IDF Israel Defense Forces IDF 以色列国防军
IED improvised explosive device IED 简易爆炸装置
LOA level of autonomy LOA 自治水平
MBC management by consent MBC 经同意管理
MBE management by exception MBE 例外管理
MIX Mix Initiative Experimental MIX 混合倡议实验
MODSIM Modeling and Simulation Environment MODSIM 建模和仿真环境
MP miss prone MP 容易出现小姐
NAO named area of interest NAO 指定的兴趣范围
27 27
ROE rules of engagement ROE 交战规则
SA situation awareness SA 情况意识
SOURCE Safe Operations for Unmanned systems for Reconnaissance in Complex SOURCE 用于复杂侦察的无人系统的安全操作
Environments Environments
SPAWAR Space and Naval Warfare Systems Center SPAWAR 空间和海战系统中心
SUGV small unmanned ground vehicle SUGV 小型无人地面车辆
TP travel planner TP 旅游策划师
UAS unmanned aerial system UAS 无人飞行系统
UCF University of Central Florida UCF 中佛罗里达大学
UGV unmanned ground vehicle UGV 无人地面车辆
UV unmanned vehicle UV 无人车辆
28 28
非也。.
NO. OF NO. OF 的非也。. 的
COPIES ORGANIZATION COPIES ORGANIZATION 复制组织复制组织
1 DEFENSE TECHNICAL 1 ARMY RSCH LABORATORY – HRED 1国防技术（PDF）信息CTR 1陆军RSCH实验室-HRED（PDF）RDRL HRM CK

(PDF) INFORMATION CTR (PDF) RDRL HRM CK J REINHART J REINHART
DTIC OCA 10125 KINGMAN RD BLDG 317 DTIC OCA 10125 KINGMAN RD BLDG 317
FT. BELVOIR VA 22060-5828 FT. BELVOIR VA 22060-5828
1 DIRECTOR 1主任(PDF)美国陆军研究实验室
(PDF) US ARMY RESEARCH LAB 1 ARMY RSCH LABORATORY – HRED 1陆军RSCH实验室-HRED（PDF）RDRL HRM A
IMAL HRA (PDF) RDRL HRM AY M BARNES IMAL HRA Y M BARNES
2520 HEALY AVE 希利大街2520号
1 DIRECTOR STE 1172 BLDG 51005 1主任(PDF)美国陆军研究实验室 STE 1172 BLDG 51005
(PDF) US ARMY RESEARCH LAB FT. HUACHUCA AZ 85613-7069 FT. HUACHUCA AZ 85613-7069
RDRL CIO LL RDRL CIO LL
1 ARMY RSCH LABORATORY – HRED 1陆军RSCH实验室-HRED（PDF）RDRL HRM AP
1 GOVT PRINTG OFC (PDF) RDRL HRM AP D UNGVARSKY 1GOVT PRINTG OFC(PDF)A M D UNGVARSKY
(PDF) A MALHOTRA POPE HALL BLDG 470 ALHOTRA 教皇堂大厦470
BCBL 806 HARRISON DR BCBL 806 HARRISON DR
1 ARMY RSCH LABORATORY – HRED FT. LEAVENWORTH KS 66027-2302 1陆军RSCH实验室-HRED（PDF）RDRL HRM C FT. LEAVENWORTH KS 66027-2302
(PDF) RDRL HRM C A DAVISON A DAVISON
320 MANSCEN LOOP STE 115 1 ARMY RSCH LABORATORY – HRED 320曼森环线STE115 1陆军RSCH实验室-HRED(PDF)RDRL HRM AT J C
FT. LEONARD WOOD MO 65473 (PDF) RDRL HRM AT J CHEN 英尺。莫伦纳德*伍德65473 HEN
12423 RESEARCH PKWY 12423 RESEARCH PKWY
1 ARMY RSCH LABORATORY – HRED ORLANDO FL 32826-3276 1陆军RSCH实验室-HRED（PDF）RDRL HRM D ORLANDO FL 32826-3276
(PDF) RDRL HRM D
T DAVIS 1 ARMY RSCH LABORATORY – HRED T DAVIS 1陆军RSCH实验室-HRED（PDF）RDRL HRM AT
BLDG 5400 RM C242 (PDF) RDRL HRM AT C KORTENHAUS BLDG 5400 RM C242 C KORTENHAUS
REDSTONE ARSENAL AL 35898-7290 12350 RESEARCH PKWY 红石兵工厂AL35898-7290 12350 RESEARCH PKWY
ORLANDO FL 32826-3276 ORLANDO FL 32826-3276
1 ARMY RSCH LABORATORY – HRED 1陆军RSCH实验室-HRED（PDF）RDRL HRS EA
(PDF) RDRL HRS EA DR V J RICE 1 ARMY RSCH LABORATORY – HRED DR V J RICE 1 陆军RSCH实验室-HRED
BLDG 4011 RM 217 (PDF) RDRL HRM CU B LUTAS-SPENCER BLDG 4011 RM 217 (PDF) RDRL HRM CU B LUTAS-SPENCER
1750 GREELEY RD 6501 E 11 MILE RD MS 284 1750 GREELEY RD 6501 E 11 MILE RD MS 284
FT. SAM HOUSTON TX 78234-5002 BLDG 200A 2ND FL RM 2104 英尺。山姆休斯顿TX78234-5002 BLDG 200A 2ND FL RM 2104
WARREN MI 48397-5000 沃伦*米48397-5000
1 ARMY RSCH LABORATORY – HRED 1陆军RSCH实验室-HRED（PDF）RDRL HRM D
(PDF) RDRL HRM DG J RUBINSTEIN 1 ARMY RSCH LABORATORY – HRED G J RUBINSTEIN 1陆军RSCH实验室-HRED(PDF)发射卓越CTR
BLDG 333 (PDF) FIRES CTR OF EXCELLENCE BLDG 333
PICATINNY ARSENAL NJ 07806-5000 FIELD ELEMENT PICATINNY ARSENAL NJ 07806-5000 字段元素
RDRL HRM AF C HERNANDEZ RDRL HRM AF C HERNANDEZ
1 ARMY RSCH LABORATORY – HRED 3040 NW AUSTIN RD RM 221 1陆军RSCH实验室-HRED（PDF）ARMC野战元 3040 NW AUSTIN RD RM 221
(PDF) ARMC FIELD ELEMENT FT. SILL OK 73503-9043 素英尺。 SILL OK拢潞73503-9043
RDRL HRM CH C BURNS RDRL HRM CH C BURNS
THIRD AVE BLDG 1467B RM 336 1 ARMY RSCH LABORATORY – HRED 第三大道大厦1467B室336室 1陆军RSCH实验室-HRED（PDF）RDRL HRM A
FT. KNOX KY 40121 (PDF) RDRL HRM AV W CULBERTSON 英尺。诺克斯KY40121 V W CULBERTSON
91012 STATION AVE 车站大街91012号
1 ARMY RSCH LABORATORY – HRED FT. HOOD TX 76544-5073 1陆军RSCH实验室-HRED（PDF）AWC野战元 FT. HOOD TX 76544-5073
(PDF) AWC FIELD ELEMENT 素
RDRL HRM DJ D DURBIN 1 ARMY RSCH LABORATORY – HRED RDRL HRM DJ D DURBIN 1陆军RSCH实验室-HRED(PDF)人类RSRCH和工
BLDG 4506 (DCD) RM 107 (PDF) HUMAN RSRCH AND ENGRNG BLDG 4506 (DCD) RM 107 程师
FT. RUCKER AL 36362-5000 DIRCTRT MCOE FIELD ELEMENT 英尺。鲁克AL36362-5000 DIRCTRT MCOE场元
RDRL HRM DW C CARSTENS RDRL HRM DW C CARSTENS
6450 WAY ST 6450 WAY ST
BLDG 2839 RM 310 BLDG 2839 RM 310
FT. BENNING GA 31905-5400 英尺。本宁GA31905-5400
29 29
非也。.
NO. OF 的
COPIES ORGANIZATION 复制组织
1 ARMY RSCH LABORATORY – HRED 1陆军RSCH实验室-HRED（PDF）RDRL HRM DE

(PDF) RDRL HRM DE A MARES A MARES
1733 PLEASONTON RD BOX 3 1733 PLEASONTON RD BOX 3
FT. BLISS TX 79916-6816 FT. BLISS TX 79916-6816
8 ARMY RSCH LABORATORY – HRED 8陆军RSCH实验室-HRED（PDF）模拟和训练

(PDF) SIMULATION AND TRAINING
TECHNOLOGY CENTER 技术中心
RDRL HRT COL M CLARKE RDRL HRT COL M克拉克
RDRL HRT I MARTINEZ RDRL HRT I MARTINEZ
RDRL HRT T R SOTTILARE RDRL HRT T R SOTTILARE
RDRL HRT B N FINKELSTEIN RDRL HRT B N FINKELSTEIN
RDRL HRT G A RODRIGUEZ RDRL HRT G A RODRIGUEZ
RDRL HRT I J HART RDRL HRT I J HART
RDRL HRT M C METEVIER RDRL HRT M C METEVIER
RDRL HRT S B PETTIT RDRL HRT S B PETTIT
12423 RESEARCH PARKWAY 研究公园道12423号
ORLANDO FL 32826 ORLANDO FL 32826
1 ARMY RSCH LABORATORY – HRED 1陆军RSCH实验室-HRED(PDF)总部USASOC

(PDF) HQ USASOC
RDRL HRM CN R SPENCER RDRL HRM CN R SPENCER
BLDG E2929 DESERT STORM DRIVE 大厦E2929沙漠风暴径
FT. BRAGG NC 28310 FT. BRAGG NC 28310
1 ARMY G1 1 ARMY G1
(PDF) DAPE MR B KNAPP (PDF) DAPE MR B KNAPP
300 ARMY PENTAGON RM 2C489 300陆军五角大楼RM2C489
WASHINGTON DC 20310-0300 WASHINGTON DC 20310-0300
13 DIR USARL 13 DIR USARL

(PDF) RDRL HR (PDF) RDRL HR
L ALLENDER L ALLENDER
P FRANASZCZUK P FRANASZCZUK
C COSENZO C COSENZO
RDRL HRM RDRL HRM
P SAVAGE-KNEPSHIELD P SAVAGE-KNEPSHIELD
RDRL HRM AL RDRL HRM AL
C PAULILLO C PAULILLO
RDRL HRM AY RDRL HRM AY
M BARNES M BARNES
RDRL HRM B RDRL HRM B
J GRYNOVICKI J GRYNOVICKI
RDRL HRM C RDRL HRM C
L GARRETT L GARRETT
RDRL HRS RDRL HRS
J LOCKETT J LOCKETT
RDRL HRS B RDRL HRS B
M LAFIANDRA M LAFIANDRA
RDRL HRS C RDRL HRS C
K MCDOWELL K MCDOWELL
RDRL HRS D RDRL HRS D
B AMREIN B AMREIN
RDRL HRS E RDRL HRS E
D HEADLEY D HEADLEY
30 30

Merged Public

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Merged Public

Uploaded by

Copyright:

Available Formats

Designing for Humans in Autonomous Systems: 在自主系统中为人类设计:

Military Applications 军事应用

ARL-TR-6782 January 2014 ARL-TR-6782 January 2014

Approved for public release; distribution is unlimited. 批准公开发布;分发是无限的.

ARL-TR-6782 January 2014 ARL-TR-6782 January 2014

Designing for Humans in Autonomous Systems: 在自主系统中为人类设计:

Florian Jentsch Florian Jentsch

Tal Oron-Gilad Tal Oron-Gilad

5c. PROGRAM ELEMENT NUMBER 5C.程序元素编号

6. AUTHOR(S) 5d. PROJECT NUMBER

7. PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES) 8. PERFORMING ORGANIZATION 7. 执行机构名称和地址美国陆军研究实验室 8. 执行组织报告编号

11. SPONSOR/MONITOR'S REPORT 11. SPONSOR/MONITOR'S REPORT

12. DISTRIBUTION/AVAILABILITY STATEMENT

List of Figures iv 数字一览表 iv

2. Autonomy and Intelligent Agents 2 2. 自治和智能代理 2

3. RoboLeader and Control of Multiple Systems 3 3. 多个系统的RoboLeader和控制 3

4. Safety and Levels of Autonomy 6 4. 安全和自主水平 6

6. Situation Understanding and Decision Support 14 6. 情况理解和决策支持 14

7. Conclusions and Future Research 19 7. 结论和未来研究 19

List of Symbols, Abbreviations, and Acronyms 27 符号、缩写和首字母缩写词列表 27

Distribution List 29 分发名单 29

2. Autonomy and Intelligent Agents 2. 自治和智能代理

3. RoboLeader and Control of Multiple Systems 3. 多个系统的RoboLeader和控制

The design implications are as follows: 设计含义如下:

Figure 3. Interaction between reliability and LOA. 图3。 可靠性和LOA之间的相互作用。

Figure 4. SPAWAR robot. 图4。 SPAWAR机器人.

*Think-A-Move is a trademark of Think-A-Move Ltd. *Think-A-Move是Think-A-Move Ltd.的商标。

*Talon is a registered trademark of QuintQ_North Amarica. *Talon是QuintQ_North Amarica的注册商标。 †哈里斯控

The design implications are as follows: 设计含义如下:

6. Situation Understanding and Decision Support 6. 情况理解和决策支持

Figure 6. Ben-Gurion University combined view display. 图6。 本-古里安大学联合视图显示。

The design implications are as follows: 设计含义如下:

2-D two dimensional 2-D 二维

3-D three dimensional 3-D 三维立体

AA adaptive automation AA 自适应自动化

ANS autonomous navigation system ANS 自主导航系统

ARL U.S. Army Research Laboratory ARL 美国陆军研究实验室

BGU Ben-Gurion University BGU Ben-Gurion University

FAP false alarm prone FAP 容易出现假警报

HHD handheld display HHD 手持式显示器

HMD helmet-mounted display HMD helmet-mounted display

HRED Human Research and Engineering Directorate HRED 人类研究及工程局

HRI human-robot interaction HRI human-robot interaction

IDF Israel Defense Forces IDF 以色列国防军

IED improvised explosive device IED 简易爆炸装置

LOA level of autonomy LOA 自治水平

MBC management by consent MBC 经同意管理

MBE management by exception MBE 例外管理

MIX Mix Initiative Experimental MIX 混合倡议实验

MODSIM Modeling and Simulation Environment MODSIM 建模和仿真环境

MP miss prone MP 容易出现小姐

NAO named area of interest NAO 指定的兴趣范围

SA situation awareness SA 情况意识

SPAWAR Space and Naval Warfare Systems Center SPAWAR 空间和海战系统中心

SUGV small unmanned ground vehicle SUGV 小型无人地面车辆

TP travel planner TP 旅游策划师

UAS unmanned aerial system UAS 无人飞行系统

UCF University of Central Florida UCF 中佛罗里达大学

UGV unmanned ground vehicle UGV 无人地面车辆

UV unmanned vehicle UV 无人车辆

COPIES ORGANIZATION COPIES ORGANIZATION 复制组织 复制组织

1 DEFENSE TECHNICAL 1 ARMY RSCH LABORATORY – HRED 1国防技术（PDF）信息CTR 1陆军RSCH实验室-HRED（PDF）RDRL HRM CK

Figure 3. Interaction between reliability and LOA. 图3。可靠性和LOA之间的相互作用。

Think-A-Move is a trademark of Think-A-Move Ltd. Think-A-Move是Think-A-Move Ltd.的商标。

Talon is a registered trademark of QuintQ_North Amarica. Talon是QuintQ_North Amarica的注册商标。 †哈里斯控

Figure 6. Ben-Gurion University combined view display. 图6。本-古里安大学联合视图显示。

COPIES ORGANIZATION COPIES ORGANIZATION 复制组织复制组织