1/4

by 2030

of THE DUBAI POLICE FORCE WILL BE robots

Last May the Dubai police force took on a new, rather unconventional, recruit: the humanoid REEM made by Barcelona-based PAL Robotics. Equipped with IBM’s artificial intelligence (AI) system Watson, REEM has a touchscreen that residents can use to report a crime right away, or alternatively to pay their fines. It can scan faces and identify a person approaching at a distance of 30 metres. Even though REEM cannot at the present time chase after or arrest a suspect, it should shortly be joined by more fellow-humanoids. Abdullah Bin Sultan, director of the centre researching into future policing in Dubai, has set a precise – and very ambitious – target: by 2030 robots should make up a quarter of the city’s police force. However, this initiative is far from unique: there is a growing trend towards using robots for police work. Another trend is the use of surveillance technologies designed to help protect our cities.

REEM, FROM PAL ROBOTICS, TO JOIN THE DUBAI POLICE FORCE

Pal Robotics - Credits: @palrobotics

Robots soon the eyes and ears of the police? And much more besides!

Robot MINE-CLEARER 

Mine-clearing vehicles which avoid putting human lives at risk have been part of the police arsenal for many years. However, progress is being made with the unprecedented levels of autonomy and sophistication of the latest models. The machines, which harness the latest developments in AI, are now capable of taking action by themselves – rather than being remotely controlled – for example monitoring a specific area, recording suspicious behaviour, and so on. The arrival of these new allies is comforting news for police forces overwhelmed by potential threats. While it might not seem very likely that we will see these machines actually arresting a suspect any time soon, they could well act as the eyes and ears of the police force in strategic areas. Using this kind of system should also in certain situations avoid policemen and women having to take risks with their own safety.

Smart city

Home Surveillance: Point system works without video

Archive September 2015

Robots have also started to be used for surveillance purposes in the private sector as well and robot guards are currently being tested all over the world.  US-based Gamma 2 Robotics has developed Ramsee, a machine resembling a small cosmonaut that does the security patrol rounds at sensitive sites.

 The K5 robot made by Silicon Valley company Knightscope has also been in the news. K5 is autonomous and can patrol on its own, surveying its surroundings. When K5 detects criminal behaviour, it alerts the authorities and/or the surveillance staff. Microsoft uses the system inter alia for surveillance at a number of car parks.

Meanwhile in France, EOS Innovation has created e-vigilante, an indoor surveillance robot designed for use in warehouses. Working autonomously, the machine makes the rounds in automated fashion. If one of its sensors detects an incident – a noise, a suspicious movement, smoke escaping from somewhere, water leaking – it warns the operator in charge of the site’s remote surveillance system, who then takes over and identifies exactly what’s happening with the help of the robot’s camera. However, in spite of successful test results, for example at ID Logistics, e-vigilante has for the moment been put back in its box, as EOS innovation went into compulsory liquidation at the beginning of the year. Nevertheless, its founder and ex-CEO David Lemaitre, who is now technical director of the robotics division of the service company Econocom, is very much hoping to relaunch the venture.

Traditional video-surveillance is rarely useful for real-time crime detection 

Law enforcement agencies have another highly valuable asset in the struggle to preserve security and combat crime: the development of smart video surveillance systems. Surveillance cameras in trains, metro stations and other public places have become widespread in many cities. However, current systems suffer from a serious weakness: they require constant monitoring by a human agent. It is known that a person’s attention falls away rapidly after twenty minutes, so we can see that there is substantial room for improvement in this type of system, as a crime or offence being committed might easily escape the agent’s attention. As a result, video surveillance is mainly used as a means of dissuasion or to shed extra light on a crime or offence, and only very rarely as a real-time crime detection tool. 

"Due to the progress made in the use of algorithms and neural networks, specialists have developed increasingly complex analytical filters which are able to display relevant alerts."

For over ten years, the concept of smart video surveillance has been coming closer, with real-time video analysis systems capable of detecting intruders, monitoring traffic flows and diagnosing suspicious behaviour so as to support the work of human police officers. More recent progress in the development of algorithms and the use of AI engines have resulted in substantial improvement in reliability and in the ability of these systems to detect problems automatically. “While it is relatively easy to detect a car driving in the wrong direction in a tunnel, this is not at all the case if a person pulls out a knife in a crowd. Between these two extreme cases, due to the progress made in the use of algorithms and neural networks, specialists have developed increasingly complex analytical filters which are able to display relevant alerts," points out Dominique Legrand, President of AN2V, the French national association for cities using video surveillance.
Shutterstock
Shutterstock

The technical challenge in this field is about eliminating false alarms. “We judge a system to be inefficient if it mistakenly ‘cries wolf’ more than ten times more often than real incidents occur”, explains Legrand. Players in this market generally specialise in a given type of detection filter for a specific situation. For instance, international ‘sensitive site protection’ specialist Evitech focuses on crowd movement. Its smart analysis technology is used in the cameras of global electronics company Thales, which every year keep four million pilgrims at the holy site of Mecca safe.

This is the kind of system that ought to help the authorities avoid the deadly stampede that took place in Mecca in 2015. Video content analysis specialist Foxstream – which specialises in intrusion detection via monitoring of area perimeters and surfaces, most often using thermal cameras – is a major player in site protection, especially at nuclear power stations and airports.

Spikenet, a French startup that was acquired by US company Brainchip in 2016, exemplifies the new generation of companies that are capitalising on the benefits of progress in the field of AI. The Toulouse-based company, which has been working for the last 15 years on neural networks and facial recognition machine learning, was one of the first firms to come up with a system able to identify faces in real time.

Real-time security management: invaluable aid for law enforcement agents

Security guards are starting to dream of a system where they are only called upon when something unusual happens

It should come as no surprise to learn that, with the advent of these robots, people are starting to worry about machines usurping the role of security guard or police officer. The proponents of robots argue that such fears are unfounded and that the most important thing is to free people up from dangerous or onerous jobs. Explains David Lemaitre: “e-vigilante’s task is to go round a warehouse on a regular schedule. This means a security guard does not have to go to particular area in the warehouse him- or herself and so risk being attacked. This kind of problem most often arises when a guard leaves his or her post to patrol the site.” Basically, we need to take proper advantage of the undeniable benefits robots have to offer: a robot is tireless, will reply patiently to the same question a thousand times, and is not prone to a lack of concentration or distractions. “Given that drowsiness is a very common problem among security guards, these are substantial benefits,” underlines Lemaitre. With robots that can detect an intrusion, pick up an abandoned parcel, or spot a person behaving in an unusual manner, surveillance guards may now be starting to dream of a system where they are only called upon when something unusual happens. And some current initiatives are getting very close to making that dream come true.

video protection

4,000 cameras have been deployed in the administrative department of Yvelines in the Greater Paris area. The system comprises a set of black screens which only light up when the software detects an anomaly. The job of the guard or police officer is just to check on what has happened. Basically, the screen will only light up when a particular ‘event’ occurs – a break-in, an abnormal sound, someone lying on the ground, a broken window, etc – in a school building, one of Yvelines administrative offices or the local fire stations. Once alerted, the surveillance centre operator will carry out a ‘doubt removal’ procedure but if s/he sees that there is really a problem, s/he will immediately contact the police or the fire and rescue service and summon them to the problem area. This venture highlights how video surveillance with state-of-the-art technology could be used in the future to enable operators to intervene in real time.

At the present time it is hard to imagine a robot replacing a human being. A robot’s capacity for analysing situations falls far short of human intuition and reasoning. Not least, robots are limited in terms of movement. Mine-clearing, reconnaissance and assistance robots – i.e. the ones that law enforcement organisations and security services prefer to use – are actually slow-moving wheeled vehicles and it is very difficult for them to surmount obstacles such as barriers, climb a staircase or simply mount the kerb. With the exception of Atlas, Boston Dynamics’ flagship robot, very few humanoid robots exist today. Managing the intricacies of bipedal motion is one of the trickiest aspects of robotics, and in any case their current price is likely to dissuade any police department from going that route.

By Olivier Discazeaux