Robotics INNOVATION REPORT
Robotics INNOVATION REPORT
Robotics INNOVATION REPORT
Robotics
October 2018
INTERPOL For official use only
INNOVATION REPORT:
Robotics in Innovative Policing
September 2018
EXECUTIVE SUMMARY
Robotics is a field of science that is becoming an ever present capability of the modern
landscape. Robots tend to possess some or all of the following abilities and functions: they are
electronically programmable, capable of data processing and maintain a perception of the
physical environment in which they evolve. They also operate autonomously to some degree
and have migrated from the factory floor in manufacturing to our homes and law enforcement
agencies. Robotics coupled with advancements in artificial intelligence through machine
learning, are getting smarter, and capable of replacing many human functions and tasks. For law
enforcement they are a tool of opportunity, whether it is in a forensic lab, handling hazardous
material or used to examine and defuse an explosive in our cities, the use and application of
robotics within our agencies will increase. Unfortunately it also provides criminals with a tool
that they can use to exploit others, and police need to understand these applications and what
can be done to circumvent this threat.
1. Building relationships with industry, academic and other partners that will:
a. Aid your agency in being aware of the technology that is available and is currently
being used within your community
b. Be conscious of what technology is emerging in the near future, and what
potential opportunities and threats that this technology presents to your
community; and
c. Build prototype capability with your partners so that your agencies can be well
positions to consider the opportunities of robotic investigations in your agency.
1. Introduction
In recent years, the field of robotics has passed through unparalleled development. An increasing
number of law enforcement agencies adopt robotics on a variety of their activities/operations
ranging from a pure communication and front desk officer purpose to explosive handling.
You will find in this paper a classification of the various families of robotics accompanied by
concrete examples in relation to policing. Each of these robotic families employs a given set of
technologies that we will define (mechanics, sensors, artificial intelligence, etc.). Concrete
examples of these technologies demonstrate how they are employed to build purpose made
robots. Defining and classifying how robots can be used in policing and what potential challenges
police forces can expect to face when adopting such technology. We will use concrete examples
of current uses of robotics in police forces to illustrate and support our proposals for a way
forward and conclusion.
2. Defining Robotics
Robotics is the branch of technology that deals with the design, construction, operation, and
application of robots. Robots tend to possess some or all of the following abilities and functions:
they are electronically programmable, capable of data processing and maintain a perception of
the physical environment in which they evolve. They also operate autonomously to some degree
(i.e. moving around, operating physical movements, etc.), they sense and manipulate their
environments, and they exhibit intelligent behaviours, including behaviours that mimic living
creatures including humans. As the field of robotics significantly varies depending on their usage,
the focus of this discussion will be on the areas of interest for police use.
Traditionally, industrial robots have been programmed with a sequence control and teaching pendants;
however, in recent years, robots have been equipped with a programming interface that allows users to
manipulate the posture of the robot directly. The most advanced industrial robots are equipped with
“Adaptive control” technology giving them some flexibility to compute the most efficient actions to
achieve a given goal.
Exoskeleton robots are primarily used to enhance the human body capabilities, lifting heavy
weights for instance. They may also be used to remotely control a slave robot that will precisely
replicate each movement of the exoskeleton. These types of machinery are used in medical fields
and have equal potential for police central scientific labs to execute actions remotely via such a
tethered robot.
Figure 4 Army robot used in Iraq to dispose of Improvised Explosive Devices (IEDs)
With the assistance of advanced artificial intelligence algorithms, Autonomous Lethal Weapon
System (LAWS) or “unmanned systems” have the ability to identify and engage enemies without
any human intervention. No country currently claims to have this technology, and a revision of
the Convention on Conventional Weapons is being discussed to include provisions to control
LAWS. Such autonomous systems are known as Human-off-the-loop. An example of infamous
and straightforward LAWS are landmines.
Despite its benefits, the deployment of robots in the military context entails the same pitfalls as
any other electronic system but with the potential for more damage: there is always a possibility
of unexpected errors or malfunctioning and the risk of the enemy hijacking/hacking the devices.
Vehicle manufacturers are quite advanced in their command of the technology for autonomous
driving. An example; in April 2016, six convoys of trucks (6 different manufactures) successfully
and autonomously followed the leading truck that was being driven across Europe from various
origins in Sweden, Denmark, and Germany to Rotterdam in The Netherlands. Known as the
European Platoon challenge, using wifi technology, the following trucks maintained speed and
distance from the lead truck. Although there were drivers on hand in each of the vehicles it
demonstrated that technology could be utilised to undertake the movement of goods across
land.
The infrastructures around the vehicles (the other motor vehicles, roads, traffic lights, police
controls, etc.) need to be integrated with the communication framework before such
autonomous cars can fully bloom. 5G communication will be one of the critical technology
facilitators for this evolution. The legal framework, however, may require much longer to evolve.
For Law Enforcement this will represent a real challenges: how will Police summon an
autonomous vehicle? How will police officers control the conformity of the vehicles' firmware,
autonomous systems and sensors? Let alone large-scale cyber-attacks on these infrastructures.
3.1. SLAM
Simultaneous Localization and Mapping (SLAM) is used in a wide range of technologies, from a
home robot vacuum cleaner learning the geometry of your house to the realisation of automatic
travelling systems found in popular apps such as Waze (a Google brand).
SLAM is a technology that combines multiple sensors, simultaneously estimating the location of
the device itself and mapping its surroundings as it is exploring them. SLAM no longer requires
prior inputs such as a preloaded map and positions. However, a large amount of calculation (and
therefore processing power) is needed. Laser range scanners (LIDARs), cameras (visual, thermal
or infrared), rotation speed encoders (if the device is on wheels), microphone arrays and other
additional sensing equipment may be utilised.
SLAM enabled robots combined with LIDARs, and 3D Crime Scene Mapping cameras may
autonomously document a crime scene. While crime scene analysts would take pictures and
notes, a SLAM would automatically and simultaneously scan and document the crime scene,
quite possibly more accurately and efficiently than humans could.
Mission Planner software is used to create a mission roadmap consisting of waypoints, speeds,
altitudes or depths (for airborne and underwater systems), headings and timed information; all
plotted on a map. The mission roadmap is then uploaded to the robot’s Auto-Pilot Module.
Currently various commercial systems are competing for this market with and Open Source
projects available that can provide the capability 1. Cheap Auto Pilot hardware supported by
Open Source software are readily available and cost less than 100 Euros to implement.
Figure 5 Low-cost Auto Pilot modules hardware for open source consumer market projects
and a Mission Planner user interface
This technology has an incredible adoption rates throughout many sectors ranging from
hobbyists, farming, military, security and safety industries.
Criminals have also been early adopters of these technologies and are deploying them in their
criminal activities (deliveries in prisons, dropping explosives, site surveillance, etc.). They use
automation to avoid exposing their pilots and due to the low cost of entry will sacrifice cheap
equipment for potential capture.
Autopilot modules and mission planning coupled with the above SLAM technology have
demonstrated a strong potential to support Policing in reducing human patrols, in analysing a
crime scene and gathering evidence or even in helping prevent crimes by proactively mapping
threat intelligence.
Neural Network hardware has drastically increased the efficiency of computers in learning how
to recognise objects in images, analyse speech characteristics and solve other complex
problems. The development of a technique called “Deep Learning” achieved further remarkable
improvements in accuracy compared to “standard” Machine Learning. Deep Learning
necessitates large-scale data sets and considerable computer resources for the training phase.
Machine Learning techniques are used in predictive policing by detecting patterns through the
compiling of historical crime series and indicators. Effectively, Machine Learning algorithms have
the ability to adapt to compile and make sense of heterogeneous datasets from diverse sources.
For instance, Machine Learning is used in predicting street drug deals based on historical police
reports, public neighborhood watch reports and the city records of issues such as the instances
of damages on street lighting. Systems linked the occurrence of damage on a street light
combined with historical reporting, triggering an automated alert for a probable drug deal
occurring, advising first responders of the probably incident. Machine Learning has been
recognised as a tool for LEAs to automatically identify criminal patterns and prevent the crimes
as far back as 20132.
The transition of machine learning principles algorithms to robotics is well underway and it is
likely we will see significant advancement in robotic capability over the coming years, similarly
to the use of machine learning and neural networks in mobile telephony and drones for facial or
obstacle recognition.
To date, it is likely that criminals will also use these technology advancements and apply them
to roots to achieve their objectives. Criminals have exploited cloud computing to date as part of
their criminal activities and it is likely that they will be earl adopters of capabilities that can
facilitate in their illicit activities. Although outside the scope of this paper investigators will face
challenges in collecting evidence not only from the physical devices but also from cloud services
and large-scale infrastructure hidden beyond a physical device such as a robot.
Utilising ROS makes it possible for robot developers to focus on the development of specific
functions. ROS and its standardisation for communication made it possible to utilise the
computational resources on the cloud rather than on the robot. ROS is commonly used in
industrial applications from automotive to healthcare and manufacturing.
Any vulnerability in such a standardised and widely deployed ROS increases the opportunities
for criminals to exploit them.
The policing community is starting to contemplate how it needs to understand and develop
capabilities to police the technologies and their application (use) into the future. The Internet-
of-Things (IoT), and Smart Cities conferences focus on these challenges and work with countries
around the world to understand the requirements for law enforcement and how best these
issues can be managed. There will be an expectation by the community that LEAs will be
prepared to examine and investigate procedures of hacked ROS with a crime-specific forensics
investigation framework for methodologies and standards for acquiring digital evidence from
such systems.3
LEAs are using robotics and automated systems in general forensic science as well as digital
forensics. The automation led robots to take away precision work or repetitive laboratory
processes from humans. From extracting DNA, handling chemical traces to laser “decapsulation”
of chips in digital forensics, forensic robots equipped with suitable tools for the handling of
evidence have benefitted police investigation and operation.
October 2016, the Islamic State was the first publicly reported kill using explosives hidden inside
a drone. Consequently, the United States Military Academy’s Combating Terrorism Center at
West Point edited a few reports, including the latest report6, addressing the supply, scale and
future threats of Drones. A Sky News report7 shows ISIL seized videos demonstrating a Driverless
Car Bomb and weapon laboratory dubbed “Jihadi technical college”.
Exploiting the self-sustaining system of robotic parts easily found in the market, terrorists,
extremist groups and criminals are expanding their technology capability towards more
innovative crimes while reducing their exposure to risks. In addition to the abuse of the
autonomous system for illicit goods smuggling and trafficking as well as terrorist attacks,
criminals have adopted the technology for voice and SMS frauds. We may expect to see artificial
‘Chatbots’ used by criminals to automate and industrialise their social engineering scams.
Few Nations have a legal system in place regarding the use of autonomous systems, artificial
intelligence and robots. However, we found 9 countries have nationally funded projects to study
the Ethical Legal and Social Implications (ELSI)8. These are driving conclusion for lawmakers and
define frameworks for research projects. Whilst originally created to address the potential risks
in Genome research projects, the ELSIs have produced guidelines for a “Responsible Research
and Innovation (RRI)”. The RRI is a guideline for projects to take into account their effects and
potential impacts on the environment and society. Horizon 2020 European Commission projects9
have to comply with the ELSA RRI framework (the ELSA is the European equivalent of the ELSI).
As much as new technologies such as robots have a potential for Law Enforcement, they
inevitably pause a concern for ELSI. This concern needs to be mitigated, possibly via public
consultation, before the implementation of these technologies within the forces.
8. References:
2 Wang, Tong, Cynthia Rudin, Dan Wagner, and Rich Sevieri, "Learning to Detect Patterns of Crime."
European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in
Databases, ECML PKDD 2013, Prague, 23-27 September 2013.
3Dieber, Bernhard & Breiling, Benjamin & Taurer, Sebastian & Kacianka, Severin & Rass, Stefan & Schartner,
Peter. (2017). Security for the Robot Operating System. Robotics and Autonomous Systems. 98.
10.1016/j.robot.2017.09.017.
4 https://www.businessinsider.com/crystal-meth-drug-drone-crash-in-mexico-2015-1/?IR=T
5 https://www.mercurynews.com/2017/07/07/south-carolina-inmate-captured-in-texas-after-second-
escape/
8 https://en.wikipedia.org/wiki/Ethical,_Legal_and_Social_Aspects_research
Fig.1 https://en.wikipedia.org/wiki/Industrial_robot
Fig.2 https://www.knightscope.com/
Fig.3 https://www.kuka.com
Fig.4 https://en.wikipedia.org/wiki/PackBot
Fig.5 http://ardupilot.org/planner/
Fig.6 https://en.wikipedia.org/wiki/Robot_Operating_System
Fig.7 https://en.wikipedia.org/wiki/Sanford_Burnham_Prebys_Medical_Discovery_Institute
Fig.8 https://www.controllaser.com FA-Cube
edgci-ic@interpol.int
WWW.INTERPOL.INT