Algorithms and Architectures for

Parallel Processing 18th International

Conference ICA3PP 2018 Guangzhou
China November 15 17 2018
Proceedings Part III Jaideep Vaidya
Visit to download the full and correct content document:
https://textbookfull.com/product/algorithms-and-architectures-for-parallel-processing-1
8th-international-conference-ica3pp-2018-guangzhou-china-november-15-17-2018-pr
oceedings-part-iii-jaideep-vaidya/
More products digital (pdf, epub, mobi) instant
download maybe you interests ...

Algorithms and Architectures for Parallel Processing

18th International Conference ICA3PP 2018 Guangzhou
China November 15 17 2018 Proceedings Part I Jaideep
Vaidya
https://textbookfull.com/product/algorithms-and-architectures-
for-parallel-processing-18th-international-conference-
ica3pp-2018-guangzhou-china-november-15-17-2018-proceedings-part-
i-jaideep-vaidya/

Algorithms and Architectures for Parallel Processing

18th International Conference ICA3PP 2018 Guangzhou
China November 15 17 2018 Proceedings Part IV Jaideep
Vaidya
https://textbookfull.com/product/algorithms-and-architectures-
for-parallel-processing-18th-international-conference-
ica3pp-2018-guangzhou-china-november-15-17-2018-proceedings-part-
iv-jaideep-vaidya/

Algorithms and Architectures for Parallel Processing

ICA3PP 2018 International Workshops Guangzhou China
November 15 17 2018 Proceedings Ting Hu

https://textbookfull.com/product/algorithms-and-architectures-
for-parallel-processing-ica3pp-2018-international-workshops-
guangzhou-china-november-15-17-2018-proceedings-ting-hu/

Algorithms and Architectures for Parallel Processing

20th International Conference ICA3PP 2020 New York City
NY USA October 2 4 2020 Proceedings Part III Meikang
Qiu
https://textbookfull.com/product/algorithms-and-architectures-
for-parallel-processing-20th-international-conference-
ica3pp-2020-new-york-city-ny-usa-october-2-4-2020-proceedings-
Algorithms and Architectures for Parallel Processing
14th International Conference ICA3PP 2014 Dalian China
August 24 27 2014 Proceedings Part II 1st Edition Xian-
He Sun
https://textbookfull.com/product/algorithms-and-architectures-
for-parallel-processing-14th-international-conference-
ica3pp-2014-dalian-china-august-24-27-2014-proceedings-part-
ii-1st-edition-xian-he-sun/

Algorithms and Architectures for Parallel Processing

14th International Conference ICA3PP 2014 Dalian China
August 24 27 2014 Proceedings Part I 1st Edition Xian-
He Sun
https://textbookfull.com/product/algorithms-and-architectures-
for-parallel-processing-14th-international-conference-
ica3pp-2014-dalian-china-august-24-27-2014-proceedings-
part-i-1st-edition-xian-he-sun/

Computational Science ICCS 2018 18th International

Conference Wuxi China June 11 13 2018 Proceedings Part
III Yong Shi

https://textbookfull.com/product/computational-science-
iccs-2018-18th-international-conference-wuxi-china-
june-11-13-2018-proceedings-part-iii-yong-shi/

Algorithms and Architectures for Parallel Processing

13th International Conference ICA3PP 2013 Vietri sul
Mare Italy December 18 20 2013 Proceedings Part I 1st
Edition Antonio Balzanella
https://textbookfull.com/product/algorithms-and-architectures-
for-parallel-processing-13th-international-conference-
ica3pp-2013-vietri-sul-mare-italy-
december-18-20-2013-proceedings-part-i-1st-edition-antonio-
balzanella/
Algorithms and Architectures for Parallel Processing
13th International Conference ICA3PP 2013 Vietri sul
Mare Italy December 18 20 2013 Proceedings Part II 1st
Edition Peter Benner
https://textbookfull.com/product/algorithms-and-architectures-
for-parallel-processing-13th-international-conference-
ica3pp-2013-vietri-sul-mare-italy-
 Jaideep Vaidya
Jin Li (Eds.)
LNCS 11336

Algorithms and Architectures

for Parallel Processing
18th International Conference, ICA3PP 2018
Guangzhou, China, November 15–17, 2018
Proceedings, Part III

123
Lecture Notes in Computer Science 11336
Commenced Publication in 1973
Founding and Former Series Editors:
Gerhard Goos, Juris Hartmanis, and Jan van Leeuwen

Editorial Board
David Hutchison
Lancaster University, Lancaster, UK
Takeo Kanade
Carnegie Mellon University, Pittsburgh, PA, USA
Josef Kittler
University of Surrey, Guildford, UK
Jon M. Kleinberg
Cornell University, Ithaca, NY, USA
Friedemann Mattern
ETH Zurich, Zurich, Switzerland
John C. Mitchell
Stanford University, Stanford, CA, USA
Moni Naor
Weizmann Institute of Science, Rehovot, Israel
C. Pandu Rangan
Indian Institute of Technology Madras, Chennai, India
Bernhard Steffen
TU Dortmund University, Dortmund, Germany
Demetri Terzopoulos
University of California, Los Angeles, CA, USA
Doug Tygar
University of California, Berkeley, CA, USA
Gerhard Weikum
Max Planck Institute for Informatics, Saarbrücken, Germany
More information about this series at http://www.springer.com/series/7407
Jaideep Vaidya Jin Li (Eds.)
•

Algorithms and Architectures

for Parallel Processing
18th International Conference, ICA3PP 2018
Guangzhou, China, November 15–17, 2018
Proceedings, Part III

123
Editors
Jaideep Vaidya Jin Li
Rutgers University Guangzhou University
Newark, NJ, USA Guangzhou, China

ISSN 0302-9743 ISSN 1611-3349 (electronic)

Lecture Notes in Computer Science
ISBN 978-3-030-05056-6 ISBN 978-3-030-05057-3 (eBook)
https://doi.org/10.1007/978-3-030-05057-3

Library of Congress Control Number: 2018962485

LNCS Sublibrary: SL1 – Theoretical Computer Science and General Issues

© Springer Nature Switzerland AG 2018

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the
material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation,
broadcasting, reproduction on microfilms or in any other physical way, and transmission or information
storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now
known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors, and the editors are safe to assume that the advice and information in this book are
believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors
give a warranty, express or implied, with respect to the material contained herein or for any errors or
omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in
published maps and institutional affiliations.

This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
 Preface

Welcome to the proceedings of the 18th International Conference on Algorithms and

Architectures for Parallel Processing (ICA3PP 2018), which was organized by
Guangzhou University and held in Guangzhou, China, during November 15–17, 2018.
ICA3PP 2018 was the 18th event in a series of conferences devoted to research on
algorithms and architectures for parallel processing. Previous iterations of the confer-
ence include ICA3PP 2017 (Helsinki, Finland, November 2017), ICA3PP 2016
(Granada, Spain, December 2016), ICA3PP 2015 (Zhangjiajie, China, November
2015), ICA3PP 2014 (Dalian, China, August 2014), ICA3PP 2013 (Vietri sul Mare,
Italy, December 2013), ICA3PP 2012 (Fukuoka, Japan, September 2012), ICA3PP
2011 (Melbourne, Australia, October 2011), ICA3PP 2010 (Busan, Korea, May 2010),
ICA3PP 2009 (Taipei, Taiwan, June 2009), ICA3PP 2008 (Cyprus, June 2008),
ICA3PP 2007 (Hangzhou, China, June 2007), ICA3PP 2005 (Melbourne, Australia,
October 2005), ICA3PP 2002 (Beijing, China, October 2002), ICA3PP 2000 (Hong
Kong, China, December 2000), ICA3PP 1997 (Melbourne, Australia, December 1997),
ICA3PP 1996 (Singapore, June 1996), and ICA3PP 1995 (Brisbane, Australia, April
1995).
ICA3PP is now recognized as the main regular event in the area of parallel algo-
rithms and architectures, which covers many dimensions including fundamental the-
oretical approaches, practical experimental projects, and commercial and industry
applications. This conference provides a forum for academics and practitioners from
countries and regions around the world to exchange ideas for improving the efficiency,
performance, reliability, security, and interoperability of computing systems and
applications.
ICA3PP 2018 attracted over 400 high-quality research papers highlighting the
foundational work that strives to push beyond the limits of existing technologies,
including experimental efforts, innovative systems, and investigations that identify
weaknesses in existing parallel processing technology. Each submission was reviewed
by at least two experts in the relevant areas, on the basis of their significance, novelty,
technical quality, presentation, and practical impact. According to the review results,
141 full papers were selected to be presented at the conference, giving an acceptance
rate of 35%. Besides, we also accepted 50 short papers and 24 workshop papers. In
addition to the paper presentations, the program of the conference included four key-
note speeches and two invited talks from esteemed scholars in the area, namely: Prof.
Xuemin (Sherman) Shen, University of Waterloo, Canada; Prof. Wenjing Lou, Virginia
Tech, USA; Prof. Witold Pedrycz, University of Alberta, Canada; Prof. Xiaohua Jia,
City University of Hong Kong, Hong Kong; Prof. Xiaofeng Chen, Xidian University,
China; Prof. Xinyi Huang, Fujian Normal University, China. We were extremely
honored to have them as the conference keynote speakers and invited speakers.
ICA3PP 2018 was made possible by the behind-the-scene effort of selfless indi-
viduals and organizations who volunteered their time and energy to ensure the success
VI Preface

of this conference. We would like to express our special appreciation to Prof. Yang
Xiang, Prof. Weijia Jia, Prof. Yi Pan, Prof. Laurence T. Yang, and Prof. Wanlei Zhou,
the Steering Committee members, for giving us the opportunity to host this prestigious
conference and for their guidance with the conference organization. We would like to
emphasize our gratitude to the general chairs, Prof. Albert Zomaya and Prof. Minyi
Guo, for their outstanding support in organizing the event. Thanks also to the publicity
chairs, Prof. Zheli Liu and Dr Weizhi Meng, for the great job in publicizing this event.
We would like to give our thanks to all the members of the Organizing Committee and
Program Committee for their efforts and support.
The ICA3PP 2018 program included two workshops, namely, the ICA3PP 2018
Workshop on Intelligent Algorithms for Large-Scale Complex Optimization Problems
and the ICA3PP 2018 Workshop on Security and Privacy in Data Processing. We
would like to express our sincere appreciation to the workshop chairs: Prof. Ting Hu,
Prof. Feng Wang, Prof. Hongwei Li and Prof. Qian Wang.
Last but not least, we would like to thank all the contributing authors and all
conference attendees, as well as the great team at Springer that assisted in producing the
conference proceedings, and the developers and maintainers of EasyChair.

November 2018 Jaideep Vaidya

Jin Li
 Organization

General Chairs
Albert Zomaya University of Sydney, Australia
Minyi Guo Shanghai Jiao Tong University, China

Program Chairs
Jaideep Vaidya Rutgers University, USA
Jin Li Guangzhou University, China

Publication Chair
Yu Wang Guangzhou University, China

Publicity Chairs
Zheli Liu Nankai University, China
Weizhi Meng Technical University of Denmark, Denmark

Steering Committee
Yang Xiang (Chair) Swinburne University of Technology, Australia
Weijia Jia Shanghai Jiaotong University, China
Yi Pan Georgia State University, USA
Laurence T. Yang St. Francis Xavier University, Canada
Wanlei Zhou Deakin University, Australia

Program Committee
Pedro Alonso Universitat Politècnica de València, Spain
Daniel Andresen Kansas State University, USA
Cosimo Anglano Universitá del Piemonte Orientale, Italy
Danilo Ardagna Politecnico di Milano, Italy
Kapil Arya Northeastern University, USA
Marcos Assuncao Inria, France
Joonsang Baek University of Wollongong, Australia
Anirban Basu KDDI Research Inc., Japan
Ladjel Bellatreche LIAS/ENSMA, France
Jorge Bernal Bernabe University of Murcia, Spain
Thomas Boenisch High-Performance Computing Center Stuttgart,
Germany
VIII Organization

George Bosilca University of Tennessee, USA

Massimo Cafaro University of Salento, Italy
Philip Carns Argonne National Laboratory, USA
Alexandra Carpen-Amarie Vienna University of Technology, Austria
Aparicio Carranza City University of New York, USA
Aniello Castiglione University of Salerno, Italy
Arcangelo Castiglione University of Salerno, Italy
Pedro Castillo University of Granada, Spain
Tzung-Shi Chen National University of Tainan, Taiwan
Kim-Kwang Raymond The University of Texas at San Antonio, USA
Choo
Mauro Conti University of Padua, Italy
Jose Alfredo Ferreira Costa Federal University, UFRN, Brazil
Raphaël Couturier University Bourgogne Franche-Comté, France
Miguel Cárdenas Montes CIEMAT, Spain
Masoud Daneshtalab Mälardalen University and Royal Institute
of Technology, Sweden
Casimer Decusatis Marist College, USA
Eugen Dedu University of Bourgogne Franche-Comté, France
Juan-Carlos Díaz-Martín University of Extremadura, Spain
Matthieu Dorier Argonne National Laboratory, USA
Avgoustinos Filippoupolitis University of Greenwich, UK
Ugo Fiore Federico II University, Italy
Franco Frattolillo University of Sannio, Italy
Marc Frincu West University of Timisoara, Romania
Jorge G. Barbosa University of Porto, Portugal
Chongzhi Gao Guangzhou University, China
Jose Daniel García University Carlos III of Madrid, Spain
Luis Javier García Villalba Universidad Complutense de Madrid, Spain
Paolo Gasti New York Institute of Technology, USA
Vladimir Getov University of Westminster, UK
Olivier Gluck Université de Lyon, France
Jing Gong KTH Royal Institute of Technology, Sweden
Amina Guermouche Telecom Sud-Paris, France
Jeff Hammond Intel, USA
Feng Hao Newcastle University, UK
Houcine Hassan Universitat Politècnica de València, Spain
Sun-Yuan Hsieh National Cheng Kung University, Taiwan
Chengyu Hu Shandong University, China
Xinyi Huang Fujian Normal University, China
Mauro Iacono University of Campania Luigi Vanvitelli, Italy
Shadi Ibrahim Inria, France
Yasuaki Ito Hiroshima University, Japan
Mathias Jacquelin Lawrence Berkeley National Laboratory, USA
Nan Jiang East China Jiaotong University, China
Lu Jiaxin Jiangxi Normal University, China
 Organization IX

Edward Jung Kennesaw State University, USA

Georgios Kambourakis University of the Aegean, Greece
Gabor Kecskemeti Liverpool John Moores University, UK
Muhammad Khurram Khan King Saud University, Saudi Arabia
Dieter Kranzlmüller Ludwig Maximilian University of Munich, Germany
Michael Kuhn University of Hamburg, Germany
Julian Kunkel German Climate Computing Center, Germany
Algirdas Lančinskas Vilnius University, Lithuania
Patrick P. C. Lee The Chinese University of Hong Kong, SAR China
Laurent Lefevre Inria, France
Hui Li University of Electronic Science and Technology
of China, China
Kenli Li Hunan University, China
Dan Liao University of Electronic Science and Technology
of China, China
Jingyu Liu Hebei University of Technology, China
Joseph Liu Monash University, Australia
Yunan Liu Jiangxi Normal University, China
Zheli Liu Nankai University, China
Jay Lofstead Sandia National Laboratories, USA
Paul Lu University of Alberta, Canada
Amit Majumdar University of California San Diego, USA
Tomas Margalef Universitat Autonoma de Barcelona, Spain
Stefano Markidis KTH Royal Institute of Technology, Sweden
Alejandro Masrur Chemnitz University of Technology, Germany
Susumu Matsumae Saga University, Japan
Raffaele Montella University of Naples Parthenope, Italy
Francesco Moscato University of Campania Luigi Vanvitelli, Italy
Bogdan Nicolae Argonne National Laboratory, Germany
Francesco Palmieri University of Salerno, Italy, Italy
Swann Perarnau Argonne National Laboratory, USA
Dana Petcu West University of Timisoara, Romania
Salvador Petit Universitat Politècnica de València, Spain
Riccardo Petrolo Rice University, USA
Florin Pop University Politehnica of Bucharest, Romania
Radu Prodan University of Klagenfurt, Austria
Zhang Qikun Beijing Institute of Technology, China
Thomas Rauber University Bayreuth, Germany
Khaled Riad Zagazig University, Egypt
Suzanne Rivoire Sonoma State University, USA
Ivan Rodero Rutgers University, USA
Romain Rouvoy University of Lille, France
Antonio Ruiz-Martínez University of Murcia, Spain
Françoise Sailhan CNAM, France
Sherif Sakr The University of New South Wales, Australia
Giandomenico Spezzano ICAR-CNR and University of Calabria, Italy
X Organization

Patricia Stolf IRIT, France

John Stone University of Illinois at Urbana-Champaign, USA
Peter Strazdins The Australian National University, Australia
Hari Subramoni The Ohio State University, USA
Gang Sun University of Science and Technology of China, China
Zhizhuo Sun Beijing Institute of Technology, China
Frederic Suter CNRS, France
Yu-An Tan Beijing Institute of Technology, China
Ming Tao Dongguan University of Technology, China
Andrei Tchernykh CICESE Research Center, Mexico
Massimo Torquati University of Pisa, Italy
Tomoaki Tsumura Nagoya Institute of Technology, Japan
Didem Unat Koç University, Turkey
Vladimir Voevodin Moscow University, Russia
Feng Wang Wuhan University, China
Hao Wang Shandong Normal University, China
Yu Wei Nankai University, China
Sheng Wen Swinbourne University of Technology, China
Jigang Wu Guangdong University of Technology, China
Roman Wyrzykowski Czestochowa University of Technology, Poland
Yu Xiao Shandong University of Technology, China
Ramin Yahyapour University of Göttingen, Germany
Fang Yan Beijing Wuzi University, China
Zheng Yan Xidian University, China
Laurence T. Yang St. Francis Xavier University, Canada
Wun-She Yap Universiti Tunku Abdul Rahman, Malaysia
 Contents – Part III

Big Data and Information Processing

TAMSA: Two-Stage Auction Mechanism for Spectrum Allocation

in Cooperative Cognitive Radio Networks . . . . . . . . . . . . . . . . . . . . . . . . . 3
Xinxiang Zhang, Jigang Wu, and Long Chen

QoS-Driven Service Matching Algorithm Based on User Requirements . . . . . 17

Mengying Guo and Xudong Yang

Research on Overload Classification Method for Bus Images

Based on Image Processing and SVM . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
Tingting Li, Yongxiong Sun, Yanhua Liang, Yujia Zhai,
and Xuan Ji

Accurate Acoustic Based Gesture Classification with Zero Start-Up Cost . . . . 44

Haojun Ai, Liangliang Han, Yifeng Wang, and Liang Liao

An Approach of Collecting Performance Anomaly Dataset

for NFV Infrastructure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
Qingfeng Du, Yu He, Tiandi Xie, Kanglin Yin, and Juan Qiu

An Axiomatization for BSP Algorithms . . . . . . . . . . . . . . . . . . . . . . . . . . . 72

Yoann Marquer and Frédéric Gava

Efficient and Secure Outsourced Linear Regression . . . . . . . . . . . . . . . . . . . 89

Haomiao Yang, Weichao He, Qixian Zhou, and Hongwei Li

New Multi-objectives Scheduling Strategies in Docker SwarmKit . . . . . . . . . 103

Tarek Menouer, Christophe Cérin, and Étienne Leclercq

Internet Performance Prediction Framework Based on PingER Dataset. . . . . . 118

Wei Zhang, Xiaofei Xing, Saqib Ali, and Guojun Wang

MS-RAID: An Energy-Saving Data Layout for CDP . . . . . . . . . . . . . . . . . . 132

Jingyu Liu, Ziyao Zhang, Lu Liu, and Xin Chai

Incentivizing Multimedia Data Acquisition for Machine Learning System . . . 142

Yiren Gu, Hang Shen, Guangwei Bai, Tianjing Wang, Hai Tong,
and Yujia Hu

Toward Performance Prediction for Multi-BSP Programs in ML . . . . . . . . . . 159

Victor Allombert, Frédéric Gava, and Julien Tesson
XII Contents – Part III

Exploiting the Table of Energy and Power Leverages . . . . . . . . . . . . . . . . . 175

Issam Raïs, Laurent Lefèvre, Anne-Cécile Orgerie,
and Anne Benoit

A Semantic Web Based Intelligent IoT Model . . . . . . . . . . . . . . . . . . . . . . 186

Chao Qu, Ming Tao, Jie Zhang, Xiaoyu Hong, and Ruifen Yuan

Accelerating CNNs Using Optimized Scheduling Strategy . . . . . . . . . . . . . . 196

Rui Xu, Sheng Ma, Wenwu Li, and Yang Guo

Data Analysis of Blended Learning in Python Programming. . . . . . . . . . . . . 209

Qian Chu, Xiaomei Yu, Yuli Jiang, and Hong Wang

APs Deployment Optimization for Indoor Fingerprint Positioning

with Adaptive Particle Swarm Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . 218
Jianhui Zhao, Jun Li, Haojun Ai, and Bo Cai

Deployment Optimization of Indoor Positioning Signal Sources

with Fireworks Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
Jianhui Zhao, Shiqi Wen, Haojun Ai, and Bo Cai

A Study of Sleep Stages Threshold Based on Multiscale Fuzzy Entropy . . . . 239

Xuexiao Shao, Bin Hu, Yalin Li, and Xiangwei Zheng

Blind Estimation Algorithm Over Fast-Fading Multipath OFDM Channels . . . 249

Jing Liu, Kun Han, Wenhua Wu, Shu Wang, and Xiao Yu

Facial Shape and Expression Transfer via Non-rigid Image Deformation . . . . 257
Huabing Zhou, Shiqiang Ren, Yong Zhou, Yuyu Kuang, Yanduo Zhang,
Wei Zhang, Tao Lu, Hanwen Chen, and Deng Chen

P-Schedule: Erasure Coding Schedule Strategy in Big Data

Storage System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 270
Chao Yin, Haitao Lv, Tongfang Li, Yan Liu, Xiaoping Qu,
and Sihao Yuan

Answer Aggregation of Crowdsourcing Employing an Improved

EM-Based Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 280
Ran Zhang, Lei Liu, Lizhen Cui, Wei He, and Hui Li

Internet of Things and Cloud Computing

A Parallel Fast Fourier Transform Algorithm for Large-Scale Signal Data

Using Apache Spark in Cloud . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 293
Cheng Yang, Weidong Bao, Xiaomin Zhu, Ji Wang, and Wenhua Xiao
 Contents – Part III XIII

Task Offloading in Edge-Clouds with Budget Constraint . . . . . . . . . . . . . . . 311

Lei He, Hongli Xu, Haibo Wang, Liusheng Huang, and Jingyi Ma

Motion Trajectory Sequence-Based Map Matching Assisted Indoor

Autonomous Mobile Robot Positioning . . . . . . . . . . . . . . . . . . . . . . . . . . . 327
Wenping Yu, Jianzhong Zhang, Jingdong Xu, and Yuwei Xu

Towards the Independent Spanning Trees in the Line Graphs of

Interconnection Networks. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 342
Baolei Cheng, Jianxi Fan, Xiaoyan Li, Guijuan Wang, Jingya Zhou,
and Yuejuan Han

POEM: Pricing Longer for Edge Computing in the Device Cloud . . . . . . . . . 355
Qiankun Yu, Jigang Wu, and Long Chen

Mobility Analysis and Response for Software-Defined Internet of Things. . . . 370

Zhiyong Zhang, Rui Wang, Xiaojun Cai, and Zhiping Jia

DStore: A Distributed Cloud Storage System Based on Smart Contracts

and Blockchain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 385
Jingting Xue, Chunxiang Xu, Yuan Zhang, and Lanhua Bai

Towards an Efficient and Real-Time Scheduling Platform for Mobile

Charging Vehicles. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 402
Qi Liu, Jinyang Li, Xiaoshan Sun, Junjie Wang, Yang Ning, Wei Zheng,
Jian Li, and Hengchang Liu

SoProtector: Securing Native C/C++ Libraries for Mobile Applications . . . . . 417

Ning Zhang, Guangquan Xu, Guozhu Meng, and Xi Zheng

CloudPT: Performance Testing for Identifying and Detecting Bottlenecks

in IaaS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 432
Ameen Alkasem, Hongwei Liu, and Decheng Zuo

Smart Grid Power Trading Based on Consortium Blockchain in Internet

of Things . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 453
Dong Zheng, Kaixin Deng, Yinghui Zhang, Jiangfan Zhao,
Xiaokun Zheng, and Xinwei Ma

Energy-Efficient Offloading in Mobile Edge Computing with Edge-Cloud

Collaboration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 460
Xin Long, Jigang Wu, and Long Chen

Quantitatively Investigating Multihop Localization Errors in Regular 2-D

Sensor Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 476
Bing Jia, Baoqi Huang, Tao Zhou, and Wuyungerile Li
XIV Contents – Part III

Optimizing WiFi AP Placement for Both Localization and Coverage . . . . . . . 489

Yu Tian, Baoqi Huang, Bing Jia, and Long Zhao

PLZMA: A Parallel Data Compression Method for Cloud Computing . . . . . . 504

Xin Wang, Lin Gan, Jingheng Xu, Jinzhe Yang, Maocai Xia,
Haohuan Fu, Xiaomeng Huang, and Guangwen Yang

A Caching-Based Parallel FP-Growth in Apache Spark . . . . . . . . . . . . . . . . 519

Zhicheng Cai, Xingyu Zhu, Yuehui Zheng, Duan Liu, and Lei Xu

Contextual-Field Supported Iterative Representation for

Face Hallucination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 534
Kangli Zeng, Tao Lu, Xiaolin Li, Yanduo Zhang, Li Peng,
and Shenming Qu

A Cancelable Multi-Biometric Template Generation Algorithm

Based on Bloom Filter. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 547
Lin You and Xun Li

Streaming ETL in Polystore Era . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 560

Nabila Berkani and Ladjel Bellatreche

Communication-Aware Prediction-Based Online Scheduling

in High-Performance Real-Time Embedded Systems . . . . . . . . . . . . . . . . . . 575
Baptiste Goupille-Lescar, Eric Lenormand, Nikos Parlavantzas,
and Christine Morin

Predicting SDC Vulnerability of Instructions Based on Random

Forests Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 593
LiPing Liu, LinLin Ci, and Wei Liu

Hybrid Cloud Architecture for Cross-Platform Interoperability

in Smart Homes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 608
Ming Tao, Chao Qu, Wenhong Wei, Bin Zhou, and Shuqiang Huang

Conflict-Free Block-with-Stride Access of 2D Storage Structure . . . . . . . . . . 618

Rui Song, Guozhao Zeng, Sheng Liu, and Haiyan Chen

Graph-Based Indoor Localization with the Fusion of PDR and RFID

Technologies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 630
Jie Wu, Minghua Zhu, Bo Xiao, and Yunzhou Qiu

UAV 3D Mobility Model Oriented to Dynamic

and Uncertain Environment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 640
Na Wang, Nan Di, Fei Dai, and Fangxin Liu
 Contents – Part III XV

Acquiring Hidden Space via Modifying Block Bitmap

for Android Devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 651
Wang Lianfang, Huang Hong, Li Yuanzhang, and Zhang Li

Interest Relevance-Based Caching Design in Content-Centric Networking . . . 661

Guozhi Zhang, Jiqiang Liu, Xiaolin Chang, and Yang Yang

Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 673

Big Data and Information Processing
 TAMSA: Two-Stage Auction Mechanism
for Spectrum Allocation in Cooperative
Cognitive Radio Networks

Xinxiang Zhang, Jigang Wu(B) , and Long Chen

Guangdong University of Technology, Guangzhou 510006, China

zxx 0504@163.com, asjgwucn@outlook.com, lonchen@mail.ustc.edu.cn

Abstract. Cooperative cognitive radio networks have been proposed to

address spectrum starvation problem and enhance the transmission rate
of mobile devices. Most works assume one user could afford the whole
spectrum and neglect the selfishness nature, which is not practical. Based
on group-buying, a two-stage auction mechanism named TAMSA is pro-
posed to guarantee the quality of service and improve the utilization
ratio of spectrum resources. TAMSA is an incentive mechanism involv-
ing the primary users (P U s) and relay nodes. TAMSA can also reduce
the cost of the secondary users (SU s) and increase utilities for both
P U s and relay nodes. In the first stage, SU s submit their budgets, val-
uations and demands for spectrum resources to relay nodes in group-
buying, relay nodes calculate revenues and determine the winning SU s.
In the second stage, we execute VCG auction between the relay nodes
and P U s, with a maximum-weighted-matching algorithm. TAMSA can
effectively allocate spectrum resources to meet the demands of SU s. We
show that TAMSA is truthful, individual rational and computational
efficient. Extensive simulation results show that TAMSA outperforms
random algorithm by 256% in terms of average utility of P U s. TAMSA
is able to improve the average utility of SU s and relay nodes significantly
up to 213% and 10 times respectively. TAMSA is further improved by
28.33% and 78.65% in terms of average utility of P U s over TASG and
TACC, respectively.

Keywords: Spectrum allocation · VCG auction

Incentive mechanism · Cooperative cognitive radio networks

1 Introduction
With the explosive growth of smart phones, wearable devices and Internet of
Things (IoT), they are demanding for higher data rates and lower latency. Spec-
trum resource is one of the most valuable resources for wireless communication
devices. However, many spectrum resources have been allocated to licensed users.
On one hand, existing un-used spectrum resources have become scarce. On the
other hand, some used spectrum resources have not been fully utilized, such as
c Springer Nature Switzerland AG 2018
J. Vaidya and J. Li (Eds.): ICA3PP 2018, LNCS 11336, pp. 3–16, 2018.
https://doi.org/10.1007/978-3-030-05057-3_1
4 X. Zhang et al.

radio and TV-channel, resulting in spectrum cavitation [1–3]. Cognitive radio is

proposed to solve the above problems to guarantee of Quality of Service (QoS) for
mobile devices and improve the utility ratio of spectrum resources. To enhance
the performance of cognitive radio networks (CRNs), cooperative cognitive radio
networks (CCRNs) was proposed [4].
In CCRNs, there are two kinds of users, one is the spectrum holder, that is,
the primary user (licensed user), denoted as P U s. The other is the secondary
user (unlicensed user), represented by SU s [5]. The mobile devices with cognitive
function can dynamically detect and utilize the idle spectrum resources. And
the CCRNs allows SU s to access the licensed spectrum occupied by P U s to
improve spectrum utilization [6,7], but SU s must not cause strong interference
to the normal communication of P U s. CCRNs can improve the utilization ratio
of spectrum resources by spectrum reuse.
Auction plays an important role in spectrum resources allocation since there
have been numerous researches on spectrum allocation using auctions [8–10].
Most prior works design single-seller and multi-buyer auctions with homogeneous
channels. In [1] and [4], authors design truthful auction for trading homogeneous
channels between a seller and multiple SU s. Besides, a distributed resource
allocation algorithm is adopted, and direct or cooperative transmission can be
selected with multiple sellers and multiple buyers [5]. Many studies assume that
P U s are willing to share their idle spectrum resources, In reality, P U s are usually
selfish, hence it is necessary to provide incentives for P U s to participate in.
Vickrey-Clarke-Groves (VCG) auction guarantees the truthfulness of the auction
process, which provides a new idea for resources allocation and can effectively
guarantee the economic returns of the participants. A McAfee based auction
mechanism is proposed, which considers the cooperative transmission of relay
nodes and ensures the maximum benefit of P U s, but it does not consider the
revenues of relay nodes [7]. In existing works [11–13], authors propose VCG-
based auction mechanism to maximize the utility of P U s and guarantee the
truthfulness. However, the objective is to maximize the amount of P U s neglect
the specific demands of SU s for spectrum resources.
In recent years, double auction [10], and combinatorial auction [11] have
been considered in spectrum resources allocation. However, most works neglect
data transmission cooperatively by relay nodes. Inspired by the popular group
buying services on the Internet, authors in [13] and [14] propose auction algo-
rithms based on group buying, which encourages SU s to be voluntarily grouped
together to acquire the spectrum resources in spectrum auctions. The group
buying algorithm can effectively reduce the payment of the SU s. In [12–14],
they equally distribute the spectrum resources to the winning SU s. Besides,
In [15], a multiple input and multiple output method is proposed in CRNs with
cooperative communication. It allows SU s to help data transmission for P U s
and obtain the opportunity to transmit data for themselves, but the mechanism
has a higher requirement of hardware configuration. In this work, we reduce the
payment of SU s with group buying. We allocate spectrum resources according
to the specific demands of the SU s.
 TAMSA: Two-Stage Auction Mechanism for Spectrum Allocation 5

In order to effectively allocate spectrum resources and encourage P U s to

share spectrum resources in auction we designed, we have to solve the follow-
ing challenges. (1) Running applications on mobile devices are heterogeneous,
so the budget and demand for each SU s are different. Besides, how to reduce
the cost of SU s is a challenge. (2) For spectrum holders and relay nodes should
be incentivized because of selfishness nature. Therefore, how to provide incen-
tives should be designed for both P U s and relay nodes. (3) Auction should be
truthful, budget balance, individual rational and computational efficient. Hence,
the auction mechanism should ensure the above properties. Different from the
previous works, we focus on investigating an incentive auction mechanism for
efficient spectrum resource allocation in CCRN. TAMSA provides an incentive
for both P U s and relay nodes to participate in auction. Besides, in the scenario,
TAMSA is based on group buying to reduce the payments of SU s, and TAMSA
allocates spectrum resources according to the specific demands of SU s.
The main contributions of this work are summarized as follows.
• To reduce the payment of SU s effectively, we propose an auction algorithm
based on group buying for the specific demands of spectrum resources. The
auction mechanism is applicable to heterogeneous networks. The economic
properties, truthfulness, budget balance, individual rationality and computa-
tional efficiency are proved.
• We design an incentive mechanism to encourage spectrum holders to share
their idle spectrum resources, and encourage relay nodes to transmit data
cooperatively.
• Numerous numerical results demonstrate that TAMSA is superior to the algo-
rithm Random by 256% in terms of average utility of P U s. The average util-
ity of relay nodes and SU s in TAMSA outperforms Random by 10 times
and 213% respectively. TAMSA is further improved by 28.33% and 78.65%
in terms of average utility of P U s over TASG and TACC, respectively.

2 System Model and Problem Formulation

In this section, we not only focus on the system model, but also formulate the
problem to be studied. And we introduce the related economic properties that
auction scheme should be followed. The basic notations as shown in Table 1.

2.1 System Model

In this paper, we consider a cognitive network with multiple primary users and
multiple secondary users. Besides, in order to improve the channel transmission
rate, we take the relay node into account. In this scenario, as in [16], we assume
all nodes stay static in a given auction period. TAMSA scheme aims to maximize
the social welfare in a spectrum auction, which also encourages both P U s and
SU s to participate in. To maximize the utilization of spectrum resources, the
incentive mechanism should properly assign the matching between the spectrum
6 X. Zhang et al.

Table 1. Notations for system model.

Notations Meaning
P Us Set of primary users
SU s Set of secondary users
Ri The ith relay node, where i ∈ [1, M ]
Si The ith group, where i ∈ [1, ni ]
sji The jth secondary user in the ith group, 1 ≤ i ≤ M, 1 ≤ j ≤ ni
dji (k) Demand of sji for kth Channel (P Uk ), 1 ≤ k ≤ M
bji (k) The bid of sji for kth Channel
vij (k) The valuation of sji for kth Channel
Ak Ask or reserve price of kth Channel
Siw Set of winning secondary users, 1 ≤ w ≤ ni
Riw Set of winning relay nodes
P Uiw Set of winning primary users
pji (k) The payment of sji for kth Channel
pc (k) The clearing price
Fi (k) Si (k) s payment for kth relay node
Pi (k) The ith relay node Ri (k) s payment for P Uk
Bi (k) The bid of the ith relay node Ri (k) for P Uk
uji The utility of sji
UP Uk The utility of P Uk
URk The utility of Rk

resources and the demands of SU s. Trading between P U s and SU s should meet

certain requirements to benefit both parties, so P U s need to be incentivized to
provide resources, and the demands of SU s should be satisfied.
The proposed network model is shown in Fig. 1, which is a hierarchical auc-
tion consisting of m P U s and ni SU s. The P U s possess M heterogeneous chan-
nels, and each primary user has a reserved price Ak , where k ∈ [1, M ], which
is the lowest price the P Ui is willing to sell the kth channel. The P U s have
different reserved prices Ak for spectrum, and we assume each relay node can
buy at most one spectrum. In the ith group Si , where i ∈ [1, M ], there are n
SU s and Si = s1i , s2i , · · · , sni , n ∈ ni . Each sji has a bid or budget bji (k) and a
valuation vij (k) for the kth channel P Uk . And in order to improve the utilization
of spectrum resources, each sji submits the demand for spectrum dji (k) to the
P Uk . The spectrum resource is allocated according to the specific demands of
the SU s.
 TAMSA: Two-Stage Auction Mechanism for Spectrum Allocation 7

Fig. 1. Auction model.

We design an incentive mechanism to improve the utilities of P U s and relay

nodes. TAMSA is a two-stage hierarchical auction, consisting of two-single round
sealed bid auctions, called stage I auction and stage II auction respectively. In
stage I, auction goes between relay nodes and the group of secondary users Si ,
and in stage II, the auction conducts between P U s and relay nodes Ri , and the
P U s sell their spectrum resources to relay nodes. The relay node Ri (k) gathers
bid and demand from the ith group Si . Then system model executes the stage
II auction. Ri (k) submits the bid Bi (k) to P Uk , and P Uk gives the reserve price
Ak , where k ∈ [1, M ], noting that Bi (k) ≥ Ak . The relay node Ri (k) determines
the winners in group Si (k) after gathering the ith group member’s bids, and the
set of winning SU s is denoted by Siw (k), where Siw (k) ⊆ Si , and the gathered bid
is Fi (k). We assume that each group pays for at most one relay node at the same
time, because one relay node serves for multiple groups might cause transmission
delay. If it wins in this auction, relay nodes Ri will allocate spectrum resources
to the Siw (k).

2.2 Problem Formulation

The system will determine the payment of winners. To achieve fairness, payments
of winners should be proportional to the workloads of the demands. The payment
of sji (k) is formulated as

pji (k) = pc (k) · dji (k), 1 ≤ i ≤ M, 1 ≤ j ≤ ni and 1 ≤ k ≤ M, (1)

8 X. Zhang et al.

where pc (k) is the clearing price. Let uji denote the utility of secondary user sji ,
for each sji ∈ Siw . Accordingly, the utility of sji is defined as
 j
j vi (k) − pji (k), if sji ∈ Siw and pji (k) ≤ bji (k)
ui = (2)
0, otherwise.
What calls for special attention is that the payment of sji (k) should not be higher
than the budget bji (k), k ∈ [1, M ], sji ∈ Siw . The relay node Ri (k) calculates the
finance Fi (k) collected from SU s. Hence the utility of relay node Ri is

Fi (k) − Pi (k), if Ri (k) ∈ Riw
URi = (3)
0, otherwise.
Where Pi (k) is the payment of relay node for P U s. In order to encourage spec-
trum holders to share spectrum resources, each P Uk has a reserved price Ak .
The payment of relay nodes Pi (k) should be higher than the reserved price Ak ,
so the utility of P Uk is defined as

Pi (k) − Ak , if P Uk ∈ P Ukw and Ri ∈ Riw
UP Uk = (4)
0, otherwise.
In this auction, the spectrum owners P Uk s allocate spectrum resources to SU s.
The speed of channel transmission is increased by the relay nodes cooperatively.

2.3 Economic Properties

In this section, we present several economic properties clearly that we would like
to achieve. In an auction, it will not be executed until the economic properties
are satisfied.
Definition 1 (Truthfulness). An auction is truthful. If it is a dominant strategy,
any participant’s utility will be maximized for the bidder’s true valuation, and
no bidder can improve its utility by misreporting its valuation. In this paper, it
implies the auction mechanism designed by us. Each sji submits true valuation to
Ri , and each relay node Ri show its true valuation to the kth primary user P Uk .
Definition 2 (Budget Balance). An auction is in budget balance for participa-
tors if total payment from buyers are greater than the total revenue of sellers.
In our mechanism, the auction is conducted in the form of group in tier I auc-
tion. We ensure the utilities of auctioneers are nonnegative. We make sure that
the payments that the relay nodes receive from the group are no less than the
amount paid to the P U s.
Definition 3 (Individual Rationality). An auction is individual rational. The
utility of each participant is nonnegative. In TAMSA scheme, the utilities of
SU s, relay nodes Ri and P U s are nonnegative. That is, uji , URi and UP Uk are
nonnegative.
Definition 4 (Computational Efficiency). An algorithm is computational effi-
cient if the mechanism can terminate in polynomial time. In our auction mech-
anism, the selection of winning SU s, the matching of P U s and relay nodes, and
the clearing price and payment can be completed in polynomial time.
 TAMSA: Two-Stage Auction Mechanism for Spectrum Allocation 9

3 Two-Stage Auction Mechanism

In this section, we propose a truthful two-stage auction framework called
TAMSA for cognitive radio networks shown in Fig. 1. TAMSA consists of two
sub-auctions, which satisfies these properties: truthfulness, budget balance, indi-
vidual rationality and computational efficiency.

3.1 Stage I Auction

In this stage, the ni secondary users are randomly divided into multiple groups.
The groups submit their bids or budgets to relay nodes separately. Relay nodes
will conduct the auction and decide the winning group members virtually. Then
relay nodes calculate the payment of each winner and determine the final win-
ners. It will allocate channels to SU s if it gets spectrum resources in tier II
auction.
We first introduce the algorithm to buy the spectrum by group and decide the
winners (GBDW), the details are as follows. Firstly, relay node Ri collects the
bid vector b1i , b2i , · · · , bni i , demand d1i , d2i , · · · , dni i and valuation vi1 , vi2 , · · · , vini
from SU s in Si as previous mentioned. We design an algorithm to calculate
the budget vector Fi (k) for P Uk . Then, relay nodes decide the winner in the
best performance ratio way and calculate the optimal unit price for each group.
The relay node Ri sells at most 1/2 time fraction to the Si for maximizing the
revenue. Inspired by the work in [16], we sort the vector of b/d in descending,
then we can get the optimal unit price for group Si , denoted as OP T (b/d),
bi
OP T (b/d) = max i , (5)
1≤i≤|b| di
where |b| denotes the length of the array, bi and di denote the ith budget and
demand separately. The detail of the algorithm is shown in Algorithm 1.
It should be noted that the clear price is extracted from the group to ensure
truthfulness. Relay nodes select the maximum integer m by OP T (b/d), and
then eliminate m SU s with smallest budget and lowest valuation. Fi (k) is the
gathered bid from those winning SU s, and the P Uk charges Ri (k) less than
Fi (k) for trading the kth channel.
In the example, we will show how Algorithm calculates the clearing price and
determines the winner. We assume that there are 5 SU s in group i, and their
budget and demands vector are as follows: b = {2, 3, 7, 6, 8}, d = {1, 2, 3, 2.5, 4},
so b/d = {2, 1.5, 2.33, 2.4, 2}, which can be obtained by Algorithm 1. We sort
b/d in descending and calculate OP T (b/d) to get the maximum m, hence we can
get m = 4 and the clearing price is pc = 8/4 = 2. Si participates in the auction
need to pay to the ith relay node is p1i = pc × d1i = 2 × 1 = 2. In the same way,
the payment of the other 4 secondary users can be calculated separately, which
is 4, 6, 5 and 8. Therefore, the winners in ith group are s1i , s3i , s4i and s5i , and the
amount collected by the ith relay node is 21.
10 X. Zhang et al.

Algorithm 1. GBDW : Group Buying and Decide Winners

Input: Sorted vector of b/d and the valuation.
Output: The revenue of relay nodes, Siw and the payment of secondary users.
1: Let 1 ≤ m ≤ ni − 1 be a bid-independent integer.
2: Search for the maximum m in b/d gets the maximum OP T (b/d).
3: pc = b m i /m.
4: Siw (k) ← ∅
5: Fi (k) ← ∅
6: for j ← 1 to ni do
7: pji (k) ← pc · dji (k), if pji (k) < bji (k) and pji (k) < vij (k)
8: if pji (k) < bji (k) and pji (k) < vij (k) then
9: Siw (k) ← Siw (k) ∪ sji (k)
10: Fi (k) ← Fi (k) + pji (k)
11: end if
12: end for
13: return Fi (k), Siw (k).

3.2 Stage II Auction

In this procedure, auction conducts between P U s and relay nodes, and relay
nodes compete for idle spectrum resources of P U s. According to previous
research, McAfee auction mechanism cannot be utilized since it only suits for
the scenario where there are homogeneous goods to trade [17]. In order to ensure
the truthfulness of auction mechanism and apply to heterogeneous networks, we
design a spectrum resource allocation algorithm SRA based on VCG auction
mechanism.
The detail of SRA is shown in Algorithm 2. We apply VCG-based auction
mechanism to maximize the social welfare, that is, the total utility of all the
participating bidders. Relay node assigns spectrum resource to the Siw when it
wins the primary user. Relay node Ri needs to pay for winning P Uk the reward
Pi , which is calculated by algorithm SRA.
We use the bid of relay node Bi (k) and the reserve price Ak to construct a
weighted complete bipartite graph, and the weight is (Bi (k) − Ak ). Maximum-
Weighted-Matching (MWM) can optimize all utility of participator in this auc-
tion. To ensure the truthfulness of auction, we apply VCG-based auction to
calculate payments of relay nodes. The details are as follows.

4 Theoretical Analysis
In this section, we prove that TMASA satisfies the truthfulness, individual ratio-
nality, budget balance and computational efficiency.
Theorem 1. TAMSA is truthful in the network.
Proof. In the following, we focus on proving the dominant strategy for SU s. For
buyer, sji (k) ∈ Si , it will submit its true bid and demand, because it reflects its
true demand for spectrum resource.
 TAMSA: Two-Stage Auction Mechanism for Spectrum Allocation 11

Algorithm 2. SRA : Spectrum Resource Allocation

Input: Bi (k), Ak , f or∀1 ≤ i ≤ ni and 1 ≤ k ≤ M .
Output: Rw ,P U w ,Pi .
1: W ← ∅, E ∗ ← ∅, Pi ← ∅//W is the edge set in the matching graph.
2: Create a weighted complete bipartite graph G = (R, P U, W, w) and the weight of
w(Ri , P Uk ) = Bi (k) − Ak if Bi (k) ≥ Ak .
3: E ∗ ← M aximum − W eighted − M atching(W ).
4: for each (Ri , P Uk ) ∈ E ∗ do
5: Rw ← Rw ∪ {Ri }, P U w ← P U w ∪ {P Uk }
6: W  ← W \(Ri , P Uk ), R \ ← R\{Ri }
7: G−i ← (R , P U, W  , w)
∗
8: E−i ← M aximum − W eighted − M atching(W  )
∗
9: Pi ← w(E−i ) − (w(E ∗ ) − w(Ri , P Uk )) + Ak
10: end for
11: return Rw ,P U w ,Pi .

For sji ∈ Si , it can improve its utility by changing its valuation and budget
from the first branch of Eq. (2). Besides, inspired by [18], the clearing price pc is
randomly generated by the optimal price ratio.
For Ri ∈ Rw , it can obtain the maximum utility max(Bi (k) − Ak ) if it gets
the spectrum resource in this auction. If Ri ∈ / Rw , it will fail during this auction
and cannot get the spectrum resource, because (Bi (k)−Ak ) < 0. If relay node Ri
submits untruthful bid, the result will not change, when Bi (k) < Fi (k). When
Bi (k) > Fi (k), the utility of relay node URi (k) = Fi (k)−Pi = Fi (k)−Bi (k) ≤ 0,
if it submits untruthful bid.
Therefore, both relay nodes and SU s cannot improve their utility by sub-
mitting untruthful bids.

Theorem 2. TAMSA is individual rational and budget balance.

Proof. For SU s, the utility of sji (k) is calculated by vij (k) − pji (k) > 0, for ∀sji ∈
Siw , and we have proved the individual rationality of SU s. Then we prove relay
nodes are also individual rational. For relay node Ri , the minimum payment price
for relay node Ri (k) is Ak for ∀Ri ∈ Riw , Bik ≤ Fi (k) and P Uk ∈ P U w . Besides,
the utility of primary user UP Uk = Fi (k) − Pi ≥ Bi (k) − Ak > 0. Therefore,
both buyers and sellers are willing to participate in the auction. They can all
gain nonnegative utility, and TAMSA mechanism is of individual rationality and
budget balance.

Theorem 3. TAMSA is computational efficient.

Proof. We now analyze the time complexity of algorithm TAMSA. In Algorithm

1, the time complexity of the sorting process is O(ni log ni ). In Algorithm 2,
it takes O(max{ni , M }3 ) time by applying the algorithm maximum-weighted-
matching. The time complexity of computing the payment is O(ni max{ni , M }3 ).
Hence, TAMSA is computational efficient.
12 X. Zhang et al.

5 Numerical Results
In this section, we evaluate the performance of TAMSA. In heterogeneous net-
work structure of we designed, this is the first incentive scheme proposed for
the specific demands of second users and there are no existing auction schemes
to compare with. Instead, We design the upper bound (Upper) and a random
algorithm (Random) for TAMSA to compare with. Meanwhile, we also simulate
algorithms TASG and TACC to compare with. The algorithm Upper uses the
bids of buyers as the payment to maximize the revenue. In TASG and TACC,
secondary users are divided into two sets randomly and selected the winning
set from other side. TASG is based on VCG mechanism, and TACC sorts the
reserve price Ak of primary users in ascending order and the budget Bi (k) of
relay nodes in descending order. The experiment tool is MATLAB, and the
results are averaged for 100 repetitions.
We consider a heterogeneous network shown in Fig. 1. We assume that the
number of P U s is M = 5, and there are 5 relay nodes to participate in this
auction, and the number of SU s ni varies from 20 to 120 with an increment
of 20. We assume that the valuation of secondary users vij (k) and budget bji (k)
are uniform distribution, and their ranges are denote as U (50, 150) and U (5, 10)
respectively. The reserve price Ak comply with U (10, 20) following [15–18].

5.1 Simulation Results

We first investigate the running time of TAMSA, and the results are shown in
Figs. 2 and 3. From Fig. 2, we can see that the running time is no more than
0.35 s even if the amount of SU s becomes large, i.e., when there are 120 SU s. For
Fig. 3, we can see algorithm Random runs fastest, since the algorithm Random
selects winning secondary users Siw randomly.

0.4
Random
0.3 TAMSA
0.35
TASG
0.3 TACC
0.35 0.25
Running time(s)

0.3 0.25
0.25
Running Time(s)

0.2
0.2 0.2
0.15
0.15
0.1 0.15

0.05 0.1
0
30 0.1
0.05
120
20 100
Nu 80
mb 10 0
er 60 0.05
of Us 10 20 30 40 50
PU 40 of S
s 0 ber
20 Num Number of SUs

Fig. 2. Running time of TAMSA (a). Fig. 3. Running time of TAMSA (b).
 TAMSA: Two-Stage Auction Mechanism for Spectrum Allocation 13

For TACC auction mechanism, the reserve price of primary user Ak is sorted
in ascending order, and the budget of relay nodes Bi (k) is in descending order to
guarantee the utility of P U s. Besides, TACC needs to match every primary user
and relay node, so algorithm runs the slowest. The running time of algorithm
TAMSA and algorithm TASG is not large, because they use maximum-weighted-
matching algorithm to complete matching between the winning P U s and relay
nodes.
Next, to validate Theorem 2 regarding individual rationality and budget
balance of TAMSA, we show the truthfulness in Fig. 4. In this auction, the
payment of relay nodes Pi (k) is not higher than the collected from SU s Fi (k),
and each winning primary user P Uiw receives a payment not less than its reserve
price Ak from the auctioneer.
From the experimental results in Fig. 4, we can see that the utility remains
nonnegative when relay nodes submit truthful bids. But when relay node submits
an untruthful bid, its utility rapidly reduce and will continue to be negative.
Figure 4 depicts the difference of utility. Relay nodes submit truthful bid, when
the bid of relay node is less than 50. When the bid is greater than 50, the
difference between truthful and untruthful bids is presented. The utility of relay
nodes and P U s are nonnegative, because the bid of relay node is less than
the collected from SU s, and the bid is greater than the reserve price of P U s,
Bi (k) ≤ Fi (k) and Bi (k) ≥ Ak , when relay nodes submit truthful bids. The
utility of relay node is negative, that is Bi (k) > Fi (k), if it submits an untruthful
bid. In summary, as seen in Fig. 4, the utility of relay nodes cannot be improved
by submitting untruthful bid.

80 9
Truthful Upper
unTruthful Random
60 8
TAMSA
TACC
7 TASG
40
Utility of relay nodes

6
Utility of PUs

20
5
0
4

−20
3

−40 2

−60 1
10 20 30 40 50 60 70 80 90 100 20 30 40 50 60 70 80 90 100
Bids of relay nodes Number of SUs

Fig. 4. Truthfulness of TAMSA. Fig. 5. Average utility of PUs with the

number of SUs.

Figure 5 shows how the utility of primary users UP Uk varies with the number
of SU s. With the increasing number of SU s, the average utility of P U s calcu-
lated by the five algorithms is gradually increasing. On average, the proposed
algorithm TAMSA in this paper has improved 256% on the utility of P U s com-
pared with the algorithm Random. TASG is about 217% better than algorithm
14 X. Zhang et al.

Random, TACC achieves about 156% utility gains than the algorithm Random
on the UP Uk . TAMSA is further improved up by 28.33% and 78.65% over TASG
and TACC in terms of average utility of P U s, respectively. That’s because both
algorithm TAMSA and TASG apply the maximum weighted matching algorithm
to match P U s and relay nodes to ensure the maximum benefit. Besides, both
TAMSA and TASG use the auction mechanism based on VCG to ensure the
truthfulness of algorithm. The difference between TASG and TAMSA is that
TAMSA selects the winning set of SU s with the optimal cost performance, and
TASG selects the winning set with the subset of SU s’s bid by another subset. The
optimal cost performance can enhance the revenue of P U s. In TACC, although
the utility of P U s can be increase, it cannot guarantee the maximization of its
earnings.

50 4.5
Upper Upper
45 Random Random
TAMSA 4 TAMSA
40 TACC TACC
TASG TASG
35 3.5
Utility of Relays

30
Utility of SUs

3
25
2.5
20

15 2

10
1.5
5

0 1
20 30 40 50 60 70 80 90 100 20 30 40 50 60 70 80 90 100
Number of SUs Number of SUs

Fig. 6. Average utility of Relay nodes Fig. 7. Average utility of SUs with the
with the number of SUs. number of SUs.

Figure 6 depicts the average utility of relay nodes with the varying number of
SU s. We can see that TAMSA outperforms Random by about 10 times averagely,
TASG and TACC are about 7 times and 6.6 times better than Random algorithm
respectively. TAMSA is further improved up by 44.59% and 64.22% over TASG
and TACC in terms of average utility of relay nodes, respectively. The reason is
that both TAMSA and TASG use the VCG auction mechanism to calculate the
payment of relay nodes Pi (k). In Algorithm 2, we see that the payment of relay
node is effectively reduced on the premise of guaranteeing the primary user’s
revenue, so the utility of relay nodes is improved.
Figure 7 shows the relationship between the average utility of SU s and the
number of SU s. The average utility of SU s in TAMSA outperforms Random by
213%, TAMSA is able to improve the average utility of SU s in TASG up to 181%,
and TACC achieves about 115% utility gain than the Random algorithm on the
utility of SU s. TAMSA is improved up by 16.99% and 85.73% over TASG and
TACC in terms of average utility of SU s, respectively. That’s because TAMSA
selects the winning set Siw in optimal cost performance. The payment of SU s is
calculated according to their specific demands, so TAMSA effectively improves
Another random document with
no related content on Scribd:
The Project Gutenberg eBook of Six little
Bunkers at farmer Joel's
This ebook is for the use of anyone anywhere in the United States
and most other parts of the world at no cost and with almost no
restrictions whatsoever. You may copy it, give it away or re-use it
under the terms of the Project Gutenberg License included with this
ebook or online at www.gutenberg.org. If you are not located in the
United States, you will have to check the laws of the country where
you are located before using this eBook.

Title: Six little Bunkers at farmer Joel's

Author: Laura Lee Hope

Illustrator: Walter S. Rogers

Release date: October 3, 2023 [eBook #71791]

Language: English

Original publication: New York: Grosset & Dunlap, 1923

Credits: Bob Taylor, David Edwards and the Online Distributed

Proofreading Team at https://www.pgdp.net

*** START OF THE PROJECT GUTENBERG EBOOK SIX LITTLE

BUNKERS AT FARMER JOEL'S ***
THE HAY SLIPPED OFF ALONG WITH THE SIX LITTLE
BUNKERS AND ADAM.
Six Little Bunkers at Farmer Joel’s. Frontispiece—(Page 152)
SIX LITTLE BUNKERS
AT FARMER JOEL’S

BY

LAURA LEE HOPE

Author of “Six Little Bunkers at Grandma Bell’s,”
“Six Little Bunkers at Mammy June’s,” “The
Bobbsey Twins Series,” “The Bunny Brown
Series,” “The Make Believe Series,” Etc.

ILLUSTRATED BY
WALTER S. ROGERS

NEW YORK
GROSSET & DUNLAP
PUBLISHERS

Made in the United States of America

 BOOKS BY LAURA LEE HOPE
12mo. Cloth. Illustrated.

THE SIX LITTLE BUNKERS SERIES

SIX LITTLE BUNKERS AT GRANDMA BELL’S
SIX LITTLE BUNKERS AT AUNT JO’S
SIX LITTLE BUNKERS AT COUSIN TOM’S
SIX LITTLE BUNKERS AT GRANDPA FORD’S
SIX LITTLE BUNKERS AT UNCLE FRED’S
SIX LITTLE BUNKERS AT CAPTAIN BEN’S
SIX LITTLE BUNKERS AT COWBOY JACK’S
SIX LITTLE BUNKERS AT MAMMY JUNE’S
SIX LITTLE BUNKERS AT FARMER JOEL’S

THE BUNNY BROWN SERIES

BUNNY BROWN AND HIS SISTER SUE
BUNNY BROWN AND HIS SISTER SUE ON GRANDPA’S
FARM
BUNNY BROWN AND HIS SISTER SUE PLAYING CIRCUS
BUNNY BROWN AND HIS SISTER SUE AT AUNT LU’S CITY
HOME
BUNNY BROWN AND HIS SISTER SUE AT CAMP REST-A-
WHILE
BUNNY BROWN AND HIS SISTER SUE IN THE BIG WOODS
BUNNY BROWN AND HIS SISTER SUE ON AN AUTO TOUR
BUNNY BROWN AND HIS SISTER SUE AND THEIR
SHETLAND PONY
BUNNY BROWN AND HIS SISTER SUE GIVING A SHOW
BUNNY BROWN AND HIS SISTER SUE AT CHRISTMAS
TREE COVE
BUNNY BROWN AND HIS SISTER SUE IN THE SUNNY
SOUTH
BUNNY BROWN AND HIS SISTER SUE KEEPING STORE
BUNNY BROWN AND HIS SISTER SUE AND THEIR TRICK
DOG
 THE BOBBSEY TWINS SERIES
(Sixteen Titles)
THE MAKE BELIEVE SERIES
(Twelve Titles)
THE OUTDOOR GIRLS SERIES
(Thirteen Titles)
GROSSET & DUNLAP, PUBLISHERS, NEW YORK

Copyright, 1923, by
GROSSET & DUNLAP

Six Little Bunkers at Farmer Joel’s

 CONTENTS
CHAPTER PAGE
I. Russ in Danger 1
II. A Load of Flowers 13
III. The Secret 24
IV. Where is Laddie? 36
V. Off to the Farm 44
VI. Something in the Straw 54
VII. At Farmer Joel’s 64
VIII. In the Hay 74
IX. When the Cows Came Home 85
X. Buzzing Bees 97
XI. Mun Bun’s Garden 106
XII. A Strawberry Shortcake 118
XIII. The Shoe-Lace Boy 128
XIV. The Shortcake Comes Back 136
XV. An Exciting Ride 147
XVI. Off on a Picnic 155
XVII. The Ice Cave 163
XVIII. A Big Splash 172
XIX. A Fight 184
XX. Yellow and White 192
XXI. A Mad Bull 201
XXII. After Wild Flowers 208
XXIII. A Mean Boy 220
XXIV. Stung 229
XXV. The Honey Tree 236
SIX LITTLE BUNKERS AT
FARMER JOEL’S
 CHAPTER I
RUSS IN DANGER

“Margy, will you look out on the porch and see if she’s there?”
“Yes, Vi, I will. But you ought to say please to me, ’cause mother
says——”
“All right then. Please look and see if she’s there,” begged Vi,
otherwise Violet Bunker. There were six of the little Bunkers. The
other four will be out presently.
Margy, who had been looking at picture books with her year-older
sister in a room off the porch, kindly dropped her book and started
for the door.
“If she’s there bring her in—please.” Violet laughed a little as she
added the last word. She remembered what Margy had started to
say about politeness.
Violet was piling up the books, for she had just thought of
something new to play, when Margy came hurrying back into the
room.
“She isn’t there!” gasped the smaller Bunker girl.
“She isn’t?” Violet fairly gasped out the words, and you could
easily tell that she was very much excited. “Are you sure, Margy?”
“No, she isn’t there, Vi! Maybe a tramp has taken her!”
“Oh!” cried Violet, in such a loud voice that Mrs. Bunker, having
heard part of the talk, came quickly from the room where she had
been sewing.
“Who’s gone?” demanded the mother of the six little Bunkers.
“Don’t tell me Mun Bun is lost again!”
Mun Bun was the youngest of the six little Bunkers. His real name
was Munroe Ford Bunker, but that was entirely too long for the little
fellow, so he was called “Mun Bun.” It was a name he had made up
for himself.
“Where is Mun Bun? Is he lost again?” asked Mrs. Bunker, starting
to take off her apron to go in search of the “little tyke,” as she often
called him, for he certainly did get into mischief very many times.
“Mun Bun isn’t lost,” answered Violet, as she hurried out on the
porch with Margy. “He’s out in the yard with Laddie, digging a hole.”
“An’ he says he’s going to dig down to China,” added Margy.
“And I just put clean bloomers on him!” sighed Mrs. Bunker. “But
who is gone?” she asked again. “It can’t be Rose or Russ—they’re
too old to be taken by a tramp!”
There, now you have heard the names of all six of the little
Bunkers, though Russ, being nearly ten, I think, wouldn’t like to be
called “little.”
“No, it isn’t Russ or Rose,” said Margy. “I saw them going down
the street. Maybe they’re going to daddy’s office to ask him for some
money to buy candy.”
“Oh, they mustn’t do that!” exclaimed Mrs. Bunker. “This is the first
of the month and daddy is very busy. They shouldn’t have gone
there. Are you sure, Margy?”
“Oh, they didn’t zactly say they were going there,” announced
Margy. “But I thought maybe——”
“You mustn’t tell things you aren’t sure of,” said her mother. “But
who is——”
“Mother, why is daddy so busy the first of the month?” asked Vi,
forgetting for the moment all about what she had sent Margy to look
for. Violet Bunker was, as her father said, “a great girl for asking
questions.” Her mother knew this, and, fearing that Vi would get
started on a list of inquiries that would take some time to answer,
Mrs. Bunker said:
“Now don’t begin that, Vi, dear. I’ll answer just this one question,
but not any more. Your father is busy the first of the month more than
at other times because tenants pay their rents then, and he collects
the rents for a large number of people. That’s one thing a real estate
dealer, like your father, does. Now, don’t ask another question!” she
commanded, for she saw that Vi was getting ready, as Russ would
say, “to spring another.”
“I wasn’t going to ask a question,” said Vi, looking a little hurt in
her feelings. “I was going to say——”
“Wait until I find out what’s happened first,” broke in Mrs. Bunker.
“Who is missing? It can’t be any of you, for you’re all present or
accounted for, as they say in the army. Who is——”
“It’s Esmeralda!” exclaimed Violet. “I had her out on the porch
playing with Margy. Then we went in to look at the picture books, and
I forgot about Esmeralda and——”
“Russ says her name ought to be Measles ’cause she’s all
spotted,” put in Margy, with a shake of her dark, tousled hair. “But it’s
only spots of dirt.”
“Come on,” demanded Vi of Margy, taking her younger sister by
the hand. “We’ve got to find Esmeralda!”
“Oh, it’s your doll!” remarked Mrs. Bunker, with a sigh of relief. “I
thought one of you children was missing. I had quite a start. It’s only
your doll. That’s different.”
“Esmeralda is my child, even if she is only a doll,” and Vi marched
away with Margy, her head held up proudly.
“Oh, my dear, I didn’t mean that you shouldn’t want to find your
missing play child,” called Mrs. Bunker quickly, for she realized that a
little girl’s feelings might be hurt by a slighting remark about even a
dirty and spotted doll. “I only meant that I was glad none of you
children was missing. I’ll help you look for Esmeralda.”
“She isn’t out on the porch. I looked,” said Margy.
“We left her there, didn’t we?” asked Vi, for sometimes there was
so much going on at the Bunker house that to remember where one
of the many dolls or other playthings was left became a task.
 “Yes, we left Esmeralda out on the porch,” agreed Margy. “But she
isn’t there now. I looked. She’s—she’s gone!”
Margy felt almost as sad over the loss as did Vi, though
Esmeralda, or “Measles,” as Russ called her, belonged particularly to
Violet.
“Do you s’pose a tramp would take my doll, Mother?” asked Violet,
for Mrs. Bunker was now walking toward the side porch with her two
little girls.
“No, my dear, I don’t believe so,” was the answer. “What would a
tramp want with a doll?”
This puzzled Vi for a moment, but she quickly had ready a reply.
“He—he might want to give her to his little girl,” Vi said.
“Tramps, as a rule, don’t have little girls,” remarked Mrs. Bunker.
“If they had they wouldn’t be tramps.”
This gave Vi a chance to ask another question. Eagerly she had it
ready.
“Why don’t tramps have little girls?” she inquired of her mother.
“Do they run away? I mean do the little girls run away?”
“No, that isn’t the reason,” and Mrs. Bunker tried not to smile at
Vi’s eagerness. “I’ll tell you about it some other time. But show me
where you left your doll,” she added, as they reached the shady side
porch. “Esmeralda certainly isn’t here,” for a look around showed no
doll in sight.
“Oh, where can she be?” gasped Vi, now on the verge of tears.
Margy, seeing how her sister was affected, was also getting ready to
weep, but just then a merry whistle was heard around the corner of
the house. It was the merry whistle of a happy boy.
“Here comes Russ!” exclaimed Violet, for she knew her oldest
brother’s habit of being tuneful. “He’ll help me look for Esmeralda.”
“Maybe he took her,” suggested Margy.
“No. If he did he wouldn’t be coming back whistling,” decided Vi.
 Russ Bunker, next to his father the “man” of the family, swung
around the path at the side of the house. Following him was Rose,
his sister, a year younger, a pretty girl, with light, fluffy hair. And, very
often, Rose had a merry song on her lips. But as Russ was now
whistling Rose could not sing. She always said Russ whistled “out of
tune,” but Russ declared it was her singing that was off key.
“Oh, Russ!” exclaimed his mother, “you didn’t go to daddy’s office
and bother him to-day, did you, when it’s the first of the month? And
he is so busy——”
“No, Mother, I wasn’t at daddy’s office,” Russ answered. “Rose
and I just went to the store for some nails. I’m making a seesaw, and
——”
“Oh, can I be on it?” begged Margy. “I love to teeter-totter! Please,
Russ, can’t I——”
“I want a ride, too!” put in Vi.
“All right! All right!” agreed Russ, with a laugh. “You can all have
rides—Mun Bun and Laddie too—as soon as I get it made. But it’s a
lot of work and it’s got to be done right and——”
Russ paused. He could see that something was wrong, as he said
afterward. Russ was a quick thinker. Also he was always making
things about the house. These were mostly things with which to play
and have a good time, though once he built a bench for his mother.
The only trouble was that he didn’t make the legs strong enough,
and when Norah O’Grady, the cook, set a tub of water on the bench
the legs caved in and there was a “mess” in the kitchen.
“Has anything happened?” asked Russ, for he could see that his
mother and his two small sisters had come out on the porch with
some special idea in mind.
“Violet’s doll is gone,” explained Mrs. Bunker. “She left it on the
porch, and she feels sad over losing it. If you know anything about it,
Russ——”
“You mean that old Measles doll?” asked the oldest Bunker boy,
laughing.
 “She hasn’t the measles at all—so there!” and Violet stamped her
foot on the porch.
“Well, she looks so—all spotted,” added Russ, with another laugh.
Then, as he saw that Violet was ready to cry and that Margy was
going to follow with tears, Russ added: “I guess I know where your
doll is. Henry Miller just told me——”
“Oh, did he take her?” cried Violet. “If he did I’ll never speak to him
again and——”
“Now, wait a minute!” advised Russ. “You girls always get so
excited! I didn’t say Henry took your doll. I just met him and he said
he saw a dog running out of our yard with something in his mouth.
Maybe it was the dog that took your doll, Violet.”
“Oh! Oh!” cried the little girl, and she was now sobbing in real
earnest.
“Oh, the dog will eat up Esmeralda!” and Margy added her tears to
those of Violet.
“I’ll go down the street and look for her,” quickly offered Russ. He
was a kind boy that way. Of course he didn’t care for dolls, and he
was anxious to start making the seesaw, nails for which he and Rose
had gone after. But Russ was willing to give up his own pleasure to
help his little sister.
“I’ll get your doll,” he said. “I guess that dog wouldn’t carry her far
after he found out she wasn’t a bone or something good to eat.”
“She—she—she’s a nice doll, anyhow, so there!” sobbed Violet.
“An’—an’ I—I want her!”
“I guess I can find her,” offered Russ. “Here, Rose, you hold the
nails.”
Russ started on a run toward the front gate. Mrs. Bunker and the
three girls followed. As yet Laddie and Mun Bun had not heard the
excitement over the missing doll, for they were still in the back yard,
“digging down to China.”
 Russ reached the gate, looked down the road in the direction
Henry Miller had told him the dog had run with something in its
mouth, and then Russ cried:
“I see her! I see your doll, Vi! The dog dropped her in the street! I’ll
get her for you.”
Russ started on the run toward a small object lying in the dust of
the road. Before Russ could reach the doll a big automobile truck
swung around the corner and came straight for poor Esmeralda.
“Oh, she’ll be run over!” screamed Violet. “My child!”
But Russ had also seen the truck and, knowing there would be
little left of the doll if one of the heavy wheels went over her, he ran a
little faster and darted directly in front of the big lumbering,
thundering automobile.
“Russ! Russ! Be careful!” called his mother.
“Look out there, youngster!” yelled the man who was driving the
truck.
On came the heavy automobile, bearing down on Russ who was
now in the middle of the street, stooping over to pick up Esmeralda.
 CHAPTER II
A LOAD OF FLOWERS

Three of the six little Bunkers—Rose, Margy and Violet—stood

grouped around their mother, looking with anxious eyes toward
Russ, who had made up his mind that he was going to get Vi’s doll
and snatch it out of danger before the big truck reached it. But, in
doing this, Russ was also in danger himself.
“Russ! Russ! Come back!” cried his mother, darting forward.
“It’s going to run right over him!” screamed Margy.
“He’ll be smashed!” and Violet covered her eyes with her hands.
“Let the old doll go!” shouted Rose.
But Russ did not heed. Straight across the street, directly in front
of the truck he ran, and toward Vi’s doll Esmeralda that was lying in
the highway, where she had been dropped by the stray dog.
The man driving the big truck, after giving one call of warning, had
ceased, and was now doing his best either to steer out of the way, so
he would not run over Russ, or else to put on the brakes. This last
was not so easy to do as the street just there was down hill and the
truck was a heavy one.
Russ reached the doll before the truck got to it. The Bunker boy
picked up Vi’s plaything and started to run out of danger, but he
slipped on a stone and down he fell in the dust of the road.
“Oh! Oh!” cried his mother. “Oh, Russ!”
Russ was down, but, as he said afterward, he was not “out.” He
rolled to one side, out of the way of the thundering big wheels of the
truck. A moment later he was on his feet, dirty and dusty, but holding
proudly aloft the doll he had rescued.