You are on page 1of 8

The EDNA Public Safety Drone: Bullet-Stopping

Leah Margaret La Salla Ayodele Odubela Giselle Espada
Inventor, Software Architect Director of Machine Learning Sensors Specialist
Austin, Texas, USA Denver, Colorado, USA Denver, Colorado, USA

Maria Camila Belduque Correa La Shana Lewis Aaron Wood

Aeronautical Engineer Director of Aerospace IT Mechanical Engineering Student
St. Luis, Missouri, USA College Station, Texas, USA
Austin, Texas, USA

Abstract— Urban gun violence in cities across the world is Columbine High School shooting by school-aged boys was
a serious issue for public safety agencies and disaster an unprecedented type of mass shooting for U.S. history; a
management organizations. This led us to the development national debate school safety, public health, and gun control
of the EDNA drone, an aerial robotics solution designed to versus Second Amendment rights. Mainstream news outlets
equip first responders in high-risk settings with lifesaving- have been providing publicly accessible APIs that track the
edge tools for situational awareness and non-lethal conflict casualties of mass shootings. Adding to these grim statistics
resolution. The EDNA is an unmanned aerial vehicle are the deaths and injuries from recent U.S. school shootings
(UAV) that delivers the patent-pending “Predictive in 2018 - notably February 14 at Marjory Stoneman Douglas
Probable Cause” technology. The EDNA drone is designed High School in Parkland, Florida and May 18 at Santa Fe
to provide automated real-time analysis to assist teams High School in Santa Fe, Texas. These catastrophic events
entering high-risk situations where gun violence may in total stole the lives of 27, and left many others wounded
occur. By leveraging machine learning, biometric sensors, and traumatized. By the time this paper is published and
and advanced materials in the field and routing feedback presented, many more tragedies will have undoubtedly
to an intuitive augmented-reality interface, the EDNA will occurred, including those that go unreported.
provide autonomous threat detection and bullet-stopping
capabilities wherever those features are needed--to groups The U.S. is known as the most violent among developed
such as Police and Sheriff's Departments, Fire nations, with a gun violence rate much higher than those of
Departments, and EMT and emergency rescue teams. Data other Western democracies. It is a country where people are
from the EDNA drone's sensors is fed to machine learning 10% more likely to be killed by a firearm than in other high-
algorithms running on the drone in real-time. Through a income nations - and where firearms kill about as many
neural network trained on past data, the EDNA is able to people as car crashes [2]. Incidents of mass violence have
detect the presence and location of firearms and explosives, happened to people of all ages in diverse locations such as at
even through walls or other obstacles. Through the use of music venues, community centers, homes, schools, places of
advanced metal foams and composite materials, the worship, business parks, hotels, and traffic stops - wherever
armored drone can even stop bullets--functionality which people may be. At the same time, gun violence is a women’s
has obvious benefits for humanitarian deployment. issue, with regard to male violence in both public and private
spheres of modern human society. It is a statistical reality that
simply having a gun in the home makes women five times
Keywords— UAV, Humanitarian, Drones, Bullet- more likely to be killed during a domestic violence situation
Stopping, ML, CV, Public Safety, Gun Violence [3].
Prevention, Disasters, Emergency Management, De-
escalation, Situational Awareness Gun violence is a devastating international problem. The
U.S. has the highest rate of gun violence in the developed
INTRODUCTION: SOCIAL AND POLITICAL CONTEXT world; rates than the U.S, as reported by the Small Arms
On April 20, 1999 a complex and highly planned Survey [4], a Swiss nonprofit affiliated with the Graduate
massacre was carried out at Columbine High School in Institute of International and Development
Littleton, Colorado USA in which 13 people were killed and Studies. Crowded global cities face continuous public safety
more than 20 others were injured. At the time, the threats from both domestic terrorism and international
terrorism, and in a variety of settings including disaster Sensors, software, and other devices initially intended for use
management. in fields like entertainment and medicine are now being
applied to humanitarian missions through aerial
With little political progress being made towards regulatory robotics. The EDNA drone performs sensor fusion through
changes in a society saturated with guns (both legal and a set of machine learning algorithms running directly on the
illegal), emergent technology presents astonishingly drone’s onboard companion computer, an NVIDIA TX2
effective ways of saving lives, right now, within existing gun augmented with an Intel Movidius and for the sake of
regulations. Furthermore, this new solution proposes to prototyping, incorporating a Raspberry Pi 3 to overcome
ensure and facilitate constructive changes such as de- compatibility issues with certain sensors not currently
escalation, risk mitigation and harm prevention, providing a supported on the NVIDIA’s ARM64 architecture. The
conduit for conflict resolution, reduced lethality, and system integrates PX4 for autopilot and numerous open
accountability in the interests of community trust and rapport source libraries. Together, this design allowed for the
with law enforcement [5]. development of the technology we describe here, for which
a permanent patent has been filed with the U.S. Patent &
Drone usage by public safety agencies in the United States Trademark Office.
has been explored since 2006. Because of their high
acquisition the Federal Aviation Administration (FAA) We decided to innovate the PPC technology into a
released Part 107 regulations in 2016 which specify quadcopter aerial robotics platform, allowing us to design a
restrictions on drone operations [6]. The navigational tool that maximizes solution delivery for the public good due
systems these conventional UAVs are usually limited to to its compact size, flight capacity, and integrated
video monitors and remote controllers, requiring the pilot to technologies. Drones can take to the air with real-time
split their attention between inspecting the drone’s reporting devices and get to places where first responders
environmental view while also focusing separately on cannot go due to high-risk or environmental barriers. The
navigating the drone’s flight by looking down on a separate EDNA drone’s materials and technology for these
remote-control console or joystick. capabilities, along with the stages of development, are
described as follows.
The autonomous drone navigation system outlined in this
paper is designed to provide threat mitigation and situational A. Platform
awareness not subject to human bias. Our unmanned aerial The original design worked with a Parrot ARDrone and
vehicle (UAV) piloting system & bullet-stopping was primarily focused on developing the advanced piloting
deployment invention, called the EDNA, integrates several system, an endeavor which was successful, but eventually
next generation technologies into one emergent lifesaving shelved due to user feedback. Among the challenges
utility solution. The EDNA utilizes augmented reality (AR), presented by the ARDrone was packet loss, characterized by
artificial intelligence (AI), machine learning (ML), sensor data that was sent correctly but fails to arrive to its
fusion, biometric feedback, and a holographic user interface destination. In the case of the ARDrone, this is a known issue
to create a neuromechanical drone tool that delivers our caused by insufficient hardware capabilities: the AR console
“Predictive Probable Cause” (PPC) technology, the does not manage its extremely limited resources particularly
situational awareness needed by public safety-first well. Due to these hardware limitations, we ended up simply
responders to prevent and stop gun violence. Designed with using an intermediary computer in lieu of a more robust
a “platform-agnostic” ideal in mind, the EDNA system can companion computer.
be integrated into a number of high-capacity commercial
UAV platforms (we explain our current preference for the Eventually, after evaluating numerous drone platforms, we
Intel Aero in detail further on). chose to simply use the Intel Aero [7], which was designed
to be friendly for active development. The Aero drone
This invention is the world’s only UAV piloting system to platform supports standard electronics expansions and ships
detect guns (real or fake, loaded or unloaded) and vital signs with an Intel RealSense R200 sensor; this is a legacy version
in the interest of de-escalation and maximizing human of the RealSense series which we replace with an Intel
survivability. As evidence of the unbiased-by-design RealSense D415 - a high-resolution depth camera that
lifesaving capabilities the EDNA utility represents, this computes 3D depth maps to optimize CV (computer vision.)
invention has been endorsed by the American College of We added an Intel Movidius VPU (vision processing unit)
Emergency Physicians, Black Lives Matter, and the Fraternal which enhances the execution of DNNs (deep neuronal
Order of Police. In addition, the EDNA is being developed networks) and CV at very low power consumption and very
to deploy in active shooter situations as an autonomous high speed [8].
flying shield between a shooter and the public (at schools and
other large venues).
Fig. 3. The Nvidia Jetson TX2
Fig. 1. Intel Aero drone. (The platform can be replaced according to
customization options). B. Sensor Suite
The EDNA’s sensor fusion is what facilitates its
anomaly detection capabilities. This is the result of an
intricate and cooperative integration of numerous partner
technologies. which allows for elegant delivery of the PPC
functionalities. Our EDNA prototype relies on three key
elements which we will introduce here:
Radio frequency (RF) tri-dimensional sensor
Infrared sensors
Long-range CV

The Vayyar “System on a Chip” (SOC) is a millimeter wave

(MMW) RF sensor called the Walabot. This delivers RF
imaging signal pre-processing via paired aerials which can
be programmatically configured, profiled, and tuned. The
pre-processing is made available via partially pre-configured
Fig. 2a. The Movidius (invented in Ireland) VPU is a key to improved NN firmware profiles that can be swapped programmatically, as
performance for the EDNA’s reporting of the objects or people it detects. b.
The Intel RealSense D415 has 2 IR (infrared) sensors in addition to its
desired. The Walabot Maker (Developer Edition) provides
camera, allowing the EDNA to perceive depth. 18 aerials (just under half the number of aerials available in
more sophisticated versions of this sensor) and includes
Image processing is improved for a better deep learning access to the device’s raw signal output, a necessity for
performance, using technology that is a subfield of machine achieving the objectives of this prototype design. In
learning - based on algorithms inspired by the structure and particular, the Vayyar Walabot provides a selection of pre-
function of the brain and aimed to train a computer to classify configured device calibrations tuned specifically, for
images, recognize speech, etc., [9] the NVIDIA Redtail example, to compensate for a barrier made of concrete or
project, which implements the TrailNet DNN. TrailNet was rebar, or to detect certain vital signs, such as breathing. It is
created for that project by the NVIDIA team to allow fully especially adept at detecting and tracking people; more
autonomous navigation (no pre-training the flight path, no sophisticated versions of this particular sensor are capable of
point mapping, not even GPS) of the drone by determining distinguish size and posture (and whether they are children
its orientation (ie, right, left and straight, and lateral offset or adults) and can maintain distinction between identified
shifted right, centered, shifted left, with respect to the path) individuals. [12]. Leveraging these capabilities was highly
[10]. The DNN is also environmentally aware, detecting instrumental in the success of the prototype.
objects and avoiding obstacles, which is autonomous This combination of capabilities is intended for use in
collision avoidance as opposed to automated collision providing critical information and situational awareness to
avoidance. The main difference between the two lies in the first responders through rubble, barricades, and other
degree of human intervention that a system requires. obstacles. By detecting and interpreting vital signs, for
Autonomous systems have the capacity to recognize, and, example (whether from an armed individual’s movements or
analyze, the conditions of the scenario in which they lack of movements), conclusions can be drawn regarding
are. The autonomous collision avoidance system “thinks” whether that individual has been injured, is unconscious, is
and takes the spontaneous actions necessary to reach a agitated or is calm. Situational awareness informs decisions
desired state, whereas automated systems are built to respond determining when to give a specific order and where to
only to the limited situations and circumstances of the perform it; with the right information at the right time
original design [11]. provided by next-generation tools such as the EDNA, we
believe those orders can address violence while preserving
human life. This simplification likewise assists in expedited
alerting of emergency medical services.

Fig. 5. The tridimensional sensor output for location in a sensor field,

radio frequency amplitude, and the presence of a target firearm clip.

The algorithm is trained on the presence of a metallic firearm

component (9mm handgun clip or 9mm handgun). To build
the predictive engine, a statistical model is created from the
reflected amplitudes of the target and the reflected
amplitudes when there is no metallic target in the sensor
field. These two sets of data are combined and a feature
added to the dataset for the presence of a firearm target,
labeled “clip-present”. The data in this feature is either 0 for
the absence of a target or 1 for the presence of a target.

After training a neural network composed of two fully

connected layers (each with sigmoid activation functions),
we were able to achieve our first prediction accuracy of 79%
which was better than we expected and will continue to
improve in ongoing development & testing. The sigmoid
function is chosen as the activation function because its range
Fig. 4 Information displayed by the Vayyar Walabot sensor’s color
maps of density. The different circles shown on the top of the image of values exists between 0 and 1. To predict the presence or
represent parts of a human body. The red dot in the bottom of the image absence of a target, we estimate the probability it is present.
shows the location of the same person in the room. More than one person Since probability values can only fall between 0 and 1, the
can be identified by the sensor. sigmoid activation function performs well. When fed the
To detect AR-15s and 9mm firearms (the most common amplitude, the trained network is able to make predictions on
handguns used in the U.S.) [13], a unique software the presence of a metal target of matching size, shape, and
programming code builds by adding classifications into a density within the sensor field.
neuronal network. We began testing with 45mm, then 40mm
and rifle rounds, then 9mm and 22mm.
The Visual Servoing Platform (ViSP) and Robot Operating
System (ROS) consume the sensor fusion input to supplant
derived coordinates to the drone’s navigational system using
offboard mode of flight. As Figure #4 shows, it points out
people - and if they are holding something the program will
also point out such an object, along with coordinates that
allow tracking.
To perform analysis, the RF data is converted to a CSV
(Comma Separated Value) file. The features exported from
the RF sensor include the X, Y, Z coordinates of the target in
the sensor reading range, and amplitude of the reflected
Fig. 6. A sigmoid activation function represented as an S-curve between 0
and 1.
The initial training was deliberately limited to the dataset of combining its attitude with intelligent incorporation of the
a .45mm semi-automatic handgun clip scanned at a distance individual’s gaze and pose.
of roughly 4-6 cm from the sensor. The sensor field was
restricted to 5° vertically and horizontally with a depth of 1- The following images show some of these early-stage
10 centimeters, with image scans being performed every 2 tracking results.
degrees, with a resolution of 1 mm. This was done to control
for the quantity of ambient data sources inevitably being
scanned by the sensor, to ensure the majority would be
representative. A second dataset was produced, and the
algorithm performs with the ability to correctly identify
likely targets.

In order to reduce the amount of noise in the tri-dimensional

sensor output, we adjust the aperture of the sensor read arena.
We are also able to adjust the reflectiveness required to filter
out non-reflective objects.

Object identification incorporating RF sensor data:

With regard for capturing highly accurate, highly precise RF

imaging data, care was paid to the impact of scan settings on Fig. 7. Trajectory tracking is determined primarily by an individual’s gaze
FPS (frames per second.) The more intensely detailed the and pose. The image depicts some of the coordinates recorded by the drone
scan over any given arena, the greater the adverse impact on related to a person’s arm movement. The drone pilot is able to see this
information by modifying leveraged functionality.
FPS. For this reason, the RF sensor’s scanning arena was
constrained to 5° for both X and Y axis with Z between 1 and
10 cm depth initially, and was first calibrated on empty air
with no objects in its sensor field to ensure a neutral baseline.
An arbitrarily chosen “reflectiveness” (amplitude) was then
selected and the data gathered. This approach allowed the
collection of enough frames to apply visualization and
analysis techniques, including canny edge detection and
Sobel derivatives while filtering the frames for highly
reflective objects.

All the devices mentioned earlier are combined in order to

determine object characteristics such as shape and proximity
to a human being, to reflect electromagnetic waves off an
object positioned behind a structure, to create a planar map
and three-dimensional perspective of. Fig. 8. This code displays the name of the objects that appear in the
camera, a pre-trained is used and then classifications of the guns we want
to record are added to the program.
Predictive trajectory tracking:
C. Armor Material
The interpretations of the RF sensor combine with the The following section is in the prototyping stage.
multiple sensor streams produced by the Intel RealSense. For
the inception of the project, we began with the relatively Situational awareness to the degree provided by the EDNA
standard approach of feeding the streams into an existing has the potential to be game-changing for law
Convolutional Neural Network (CNN) selected for enforcement. In order to be a game-changer for everyone
performance, accuracy, and efficiency (YOLO, from else, the EDNA should be able to both act upon its hyper-
darknet.) We then simply transfer trained it as necessary. We specific conclusions and actively intervene to forcibly de-
are able to leverage existing functionality built into the escalate potential gun violence. In order to give the EDNA
sensor SDK, adding to or simply modifying respectively to UAV a bulletproof body, different materials need to be
further enhance our results. This yielded a rudimentary combined into a single armor. The U.S. National Institute of
algorithm capable of assessing and predicting the trajectory Justice (the research, development and evaluation agency of
path of a ballistic projectile fired from an identified the Department of Justice) sets the requirements of properties
“anomalous threat” (an identified gun held in a hand) by that determine the ballistic resistance of body armor. Each of
the following materials meet requirements needed by a
drone: they are highly resistant, lightweight, and durable. Fig. 10. Aluminum metallic foams with different porosities.
Integrated, they work as a whole system to stop bullets and In order to process the armor as the desired prototype shown
pulverize them on impact. in Fig. 9, a mold was built out of plaster. However,
variations in temperature led to volumetric contractions
Aluminum metallic foams (AMF), composite Kevlar and which caused it to shrink and become brittle. So, we
polyethylene fibers (both used in bulletproof vests) [14] [15], proceeded by choosing a refractory cement. After observing
are the options currently under research to manufacture a the behavior of several combinations of water, cement and
“sandwich structure” in which the core material or matrix is sand, we chose the one that better behaves under temperature
the AMF and both back and/or front plate are made out of and brittleness. Fig 11 shows the new mold (the results will
Kevlar (KV, polyparaphenylene terephthalamide), or be presented in another paper).
emergent bulletproof polyethylene fibers. The AMF utilizes
a new cellular material composed of an aluminum (Al) metal
matrix which builds a web of cells interconnected with each
other. This kind of structure exhibits a low density of
approximately 0.9g/cm3 due to its high porosity, against a
2.7g/cm3 of bulk aluminum, resulting in a significant weight
reduction and high levels of energy absorption. Other
appealing properties include good permeability, high heat
dissipation, and additional acoustic behavior. One step in the
design process for the armor prototype aimed to protect the
battery, Fig. 9 shows the actual design made in Solid Edge
ST8. Fig. 11. Refractory cement mold.

Similar experiments have taken place showing significant

results at the University of North Carolina using a stainless
steel (S-S) bulk material. [17] With a density of about
8g/cm3, the S-S is processed into a composite metallic foam
(CMF) of 3.4g/cm3, a higher strength-to-density ratio and
higher plateau strength than other metal foams. This
combination of S-S composite metallic foam, with KV or Al
backplates and a Boron Carbide B4C front plate, has been
Fig. 9. Armor prototype, designed to cover the battery. Will also be tested
proven to successfully stop NIJ type III (0.357 SIG and 0.44
ballistically to determine the drone’s performance and make the necessary Magnum) impacts at different speeds. The S-S CMF
corrections to the code in order to readjust the right balance position after an material absorbs most of the kinetic energy and provides an
impact. An opening is made on the top since this part of the process will be an additional outstanding armoring option for the EDNA.
carried out before taking off the antenna and replacing the computer board
After researching armoring materials and ballistics testing
which the Aero comes with.
results, we decided to start developing an AMF (Aluminum
Metallic Foam) armor with a likewise high capacity to
The Al alloy used as the core for the AMF is A356. The dissipate kinetic energy at a lighter weight.
pores are created by infiltrating NaCl (common salt) particles
under vacuum pressure. Both the alloy and the NaCl have
This high weight-resistance ratio material, combined with a
the added advantages of being highly available, low cost, and
machine-learning ability to track the output end of any
recyclable [16]. Some of the samples obtained through this firearm, equip the EDNA with the capacity to take bullets (up
process are shown in Fig. 10. to 0.50 caliber, pulverized on impact) and to stop shooters by
autonomously diverting full attention away from possible
victims. The raw, untreated armor dissipates most of the
incoming bullet force (even when the armor technically fails,
it simply deforms - it does not splinter or shatter), as has been
tested and demonstrated. Further testing has shown that a
fabrication of a KV or Al backplate with the CMF combined
with a front layer of B4C ceramic ensures an armor that is
absolutely bullet- and blast-resistant, to further eliminate
bullet lethality.
D. Piloting System
Early attempts:
A sophisticated piloting mechanism was originally is already aware and has adjusted their response to
developed involving an Emotiv Insight EEG sensor and a be non-lethal.
Microsoft HoloLens, which were mapped together via the 2) A real gun present y/n? y = anomaly. Any object
HoloLens to override the drone’s navigational controls and that is positively identified as a real firearm is
manual piloting interface (the hand-held RC controller.) automatically registered as an anomaly. The officer
However, while this technology proved relatively then responds to the situation according to agency
straightforward in terms of implementation, feedback from procedures, without the possibility of being
early focus groups suggested that it was simply too advanced ambushed or surprised with unanticipated lethality.
for the majority of users. As such, it was tabled from the 3) Identification of an anomaly initializes guardian
overall EDNA project, where it remains. angel mode of all identified anomalous objects. The
EDNA drone immediately attempts to assess the
Current system: degree of anomalousness of an identified weapon
while officer is in the safety of the vehicle.
The drone is designed to be semi-autonomous, meaning it 4) As long as an anomaly is present, it is tracked, and
can be navigated manually via the standard RC controller, the EDNA drone continuously performs risk
and its autonomous behavior is designed to be overridden at assessment. Risk is assessed according to proximity
by a human operator via the RC controller. By default, it is of any identified hand to any anomaly (gun.) If a
designed to autonomously respond to certain specific gun is determined to be in the hand, the drone
criteria: when it detects an anomaly, it will shift into immediately shifts to its active shooter protocol.
“guardian angel mode,” and “active shooter mode,” which is 5) Once the atcive shooter protocol is activated, the
triggered when a hand gets too close to a gun. This provides EDNA autonomously attempts to position itself
an intuitive piloting system, one that allows articulate directly in harm’s way. It will actively seek to
operation with an ultra-fast, minimal learning curve for reposition itself so that it remains in the line of fire.
operator training. The EDNA drone’s machine learning reward
system is such that it wins when it gets shot.
Currently, a Brother AirScouter provides much less invasive
AR “drone’s eye view” visual display of the real time or near
real-time sensor output. The holographic visor permits the Use Case II
pilot to interact in real-time with 3D sensors and software Location: School or large public venue firearm threat
that displays situational and environmental conditions from User: Uniformed school police officer, school robotics club
a safe distance. Therefore, dangerous tasks in potentially Deployment: The EDNA is designed to autonomously patrol
hostile or inaccessible environments are performed with no a perimeter or specific entry points of a school.
human risk.
1) Active shooter protocol.
E. Examples of EDNA Deployment: Various Scenarios 2) EDNA drones are deployed autonomously, as
Anomaly? y/n onsite security follows best practices for emergency
Y? The EDNA enters Anomaly Tracking Mode management notifications to police. The EDNAs
N? No reaction from the EDNA. get very close to the shooter (effective even on the
Use Case I other side of a wall or window), track the firearm
Location: Routine traffic stop, moving violation output end, take bullets, and closely distract the
User: Squad car LEOs (law enforcement officers) shooter until either their ammunition runs out,
Deployment: After pulling over a vehicle - and parking the surrender occurs, or an arrest is made thanks to the
squad car at an increased distance away - the pilot officer drone distracting the shooter and giving officers the
puts on their easily accessible headgear and within minutes opportunity to disarm and restrain without resorting
is ready to deploy the EDNA for situational awareness. The to lethal force.
pilot officer’s EDNA reports the license plate, and unbiased 3) Unknown threat. Best security practices are
anomalies, i.e.: implemented with an EDNA deployment to
determine y/n anomalies such as the presence of a
firearm. Best practices for forced de-escalation and
1) A real gun present y/n? n = no reaction. The pilot
public safety ensue with anomalies. This also
officer can safely proceed to approach the car,
provides the authorities with high resolution
knowing that there are toy guns or no guns. A drone
imaging of said assailant. If no anomaly is
assessment that does not identify the presence of an
identified, the EDNA will continue to patrol the
anomaly provides the law enforcement officer with
school to ensure the safety and mental wellness of
awareness that there is no possibility of lethality in
the students.
the event of escalation, and in the event that an
escalation requires the officer use force, the officer
Use Case III Ceallaigh for diligent manuscript editing. And we’d also
Location: Natural disaster zones during flooding or lava like to extend our thanks to Alexis Sachin, Victoria Cloud
flows, or after earthquakes, mudslides, tornadoes. and Nina Alter which helped us in successfully complete
User: Emergency services personnel (police, fire rescue) this document.
Deployment: The EDNA can identify vital signs of survivors
trapped in debris, mud, flooded structures and other
situations unseen to the human eye. Thus, the EDNA vastly [1] G. Lopez, (May 2018). “America’s Unique Gun Violence Problem,
Explained in 17 Maps and Charts”. United States. Retrieved from
expedites precision rescue strategie and can do so in the
presence of anomalies like loaded gun threats interfering [2] A. Boult (February 2016). “Americans are ten times more likely to die
with humanitarian relief to disaster-stricken zones. in gun deaths' than those in other countries” United States. Retrieved
Use Case IV [3] C. Campbell, D. Webster, J. Koziol-McLain, C. Block, D. Campbell,
M. Ann Curry, F. Gary, N. Glass, J. McFarlane, C. Sachs, P. Sharps,
Location: SWAT hostage, domestic firearm standoff with Y. Ulrich, A. Wilt, J. Manganello, X. Xu, J. Schollenberger, V. Frye,
children present in the vicinity. K. Laughon, “Risk Factors for Femicide in Abusive Relationships:
Results from a Multisite Case Control Study” American Public Health
Association. United States, July 2003.
Deployment: The EDNA drone is designed to deploy
[4] C. McEvoy, G Hideg, ‘‘Global Violent Deaths 2017, Time to Decide’’
autonomously to ascertain presence of any anomalies and The Small Arms Survey and Swiss Agency for Development and
whether or not this is an authentic emergency. Cooperation. Switzerland, December 2017. pp 19-30.
[5] D. Cohen, (August 2014). “Unmanned Aircraft Systems (UAS)
1) A real gun present y/n? n = no reaction. Law Guidebook in Development” Retrieved from
[6] D. Gettinger, “Public Safety Drones”, Center for the Study of the
enforcement determines this is not a real SWAT Drone at Bard College. United States, April 2017.
situation as the EDNA does not react to anything [7] Intel Software (2016, November 29). Introduccion to the intel aero
other than a firearm, bomb, or an active compute board and vision accessory kit for UAVs [Video file].
shooter. With this awareness, the SWAT unit can Retrieved from
exit the vicinity and be available for other potential e
emergencies, while peace officers proceed to [8] P. Guermonprez (November 2017) “01 About Intel Aero”. Retrieved
document the victims of the “SWATting” crime from
[18]. [9] I. Sutskever, J. Martens, J. Dahl, G. Hinton. “Of the importance of
initialization and momentum of deep learning” Department of
2) A real gun present y/n? y = anomaly. SWAT Computer Science, University of Toronto. Toronto, Canada.
protocols proceed, with law enforcement able to [10] B. Nemire, (September 2017). “NVIDIA researchers release
actively intervene despite being blocked by a wall. trailblazing deep learning-based framework for autonomous drone
The EDNA provides structural mapping, and navigation’’ Retrieved from
identifies locations of all individuals in the building [11] U. Krogmann, “From automation to autonomy, trends towards
autonomous combat systems” Germany,
during forced de-escalation and hostage rescue.
[12] Walabot 3D imaging sensor, April 2016. Vayyar Imaging Ltd.
Retrieved from
F. Conclusion [13] B. Segal, (June 2018). “These are the guns criminals in America love”.
The EDNA drone invention behavior was designed to be Retrieved from
indiscriminate; a community-based public safety solution [14] L. Alil, C. Barbu, S. Badea, F. Ilie. “Aspects regarding the use of
polyethylene fibers for personal armor” Military Technical Academy,
excellence of technological integrations in mind. Since Scientific Research Center for CBRN Defense and Ecology, Land
capturing our first neuromechanical drone piloting system in Forces Academy. Romania 2015.
2015 with a body of nine inventions, we’ve conducted [15] P. Fernandez. “Aluminum foams for functional and structural
numerous cycles of further research, development, and applications”. Universidad Pontificia Bolivariana. Colombia,
testing in active collaboration and cooperation with September 2013.
uniformed police officers, firefighters, and community [16] P. Fernandez, L. Garcia-Cambronero, L. “Use of recycled
aluminum for low cost production of metallic foams.”Universidad
leaders. As our partner technologies continue to develop, we Pontificia Bolivariana. Universidad Politécnica de Madrid, Enero
continue to adapt and improve in preparation for bringing 2010
this astonishing humanitarian utility to scale. Inventor La [17] J. Marx, M. Portanova, A. Rabiei. “A study on blast and fragment
Salla and team have demonstrated the EDNA’s prototypical resistance of composite metal foams through experimental and
capabilities to public safety agencies in the U.S, by request. modeling approaches.” North Carolina State University, Amy
Research, Development and Engineering center. USA, March 2018.
Multi-location public safety pilot programs of the EDNA
[18] X. Wang, “Augmented Reality: A new way of augmented learning”
begin in Spring 2019.
Australia, eLearn Magazine, October 2012.
[19] L. La Salla. “Biometric feedback and augmented reality device and
ACKNOWLEDGMENT system for operating an unmanned vehicle” Unpublished, in a filled
The authors acknowledge the late Joseph P. La Salla, permanent patent pending issuance.
Aviation Engineer (deceased), for foundational technical
assistance and mentorship. We also wish to thank Mary