Can algorithm save next victim of crime?

Domestic violence, theft, brawls, missing children, traffic accidents, suicide attempts, murder – police patrols in all cities around the world are called upon to respond to such events 24/7, quickly and efficiently, to save lives, prevent damage and injury, and lend a helping hand. Their work is vital, but are they really as fast and effective as they can be? 

Dr. Sofia Amador of the Faculty of Industrial Engineering and Technology Management proposes a novel technology that can really save lives – by sending forces to the rescue much more quickly and efficiently.  The task-allocation algorithm, developed by Sofia in collaboration with researchers from Ben-Gurion University and the Beer-Sheva Police Department, is able to dispatch the right patrol teams to every emergency in real time, based on parameters such as present location (distance from the scene), the urgency of each event and the professional expertise of individual officers. Moreover, Sofia emphasizes that even though the innovative algorithm is based upon police force data and needs, it can easily be adapted to serve other rescue forces, such as fire-fighting crews and ambulance fleets, or any other operation in which effective task-allocation is crucial.

"Today police operations in Israeli cities are monitored by central operation centers, from which officers allocate incoming tasks to police vehicles on patrol at the time," says Sofia. "This method is reasonably efficient when there aren't too many tasks to allocate, but when the pace of events is accelerated, and the calls come pouring in, it is far from optimal. The unaided human brain cannot process and prioritize so many items effectively, and decision making is compromised. When this happens, complexity and confusion lead to mishaps, mishandling, and in some cases – to tragedies that could and should have been prevented."

The novel solution proposed by Sofia and her colleagues is an algorithm based on an economic model known as the Fisher Market Clearing Mechanism. "It's a mechanism of buyers and sellers," explains Sofia. "Every seller sells one product and every buyer has personal preferences regarding the products, and the same sum of money to spend. Ultimately the mechanism distributes all products among all buyers, with no money left over, leaving everyone happy. No one envies anyone else (envy freeness) and the distribution of products is optimal (pareto optimal), precisely in accordance with the buyers' preferences."

Adapting this mechanism to the needs of the police, the developers defined the police patrols as buyers, and their tasks as goods to be optimally distributed.  They created a simulation of the city of Beer Sheva and its patrolling police cars, with calls coming in at various intervals – all based on real police data. To this simulation they applied two different algorithms: one, called a 'greedy' algorithm, bases its decisions on relatively simple rules of thumb - much like the human brain, especially under pressure; the other was their own newly developed algorithm.  Comparing the results of the two algorithms, the researchers found a gap in favor of their own algorithm, which grew larger as the challenge grew more complex – most evidently when the pace of incoming calls grew faster and faster. 

"Our algorithm was clearly better in all parameters," says Sofia. "Taking both time and space constraints into consideration, it provided a quicker response and faster arrival on the scene, a better response to urgent calls, fewer events left unaddressed, and fewer cases in which forces had to abandon the scene to respond to another call. Since the algorithm is very versatile, it can also be adapted to many other operations and services. At present we are working with Elbit to create a similar algorithm for decentralized autonomous systems, such as drone fleets. Other studies currently underway address possible problems in such operations – such as communication failures and the disappearance of a participating vehicle, requiring a reallocation of tasks (with Ben-Gurion University)." 

Posted: 2/03/2021