The neural network will not work according to the rules

The city authorities will spend 285 million rubles. to create a system that will use a neural network to identify various incidents and emergency situations in the road network. We are talking, for example, about the intrusion of cars and pedestrians into prohibited areas or about identifying cars moving significantly slower than the main stream. The system will also give a signal in the event of a fire or detection of foreign objects on the roadway. The Moscow City Hall promises to use PROEvent to transmit data to emergency services in order to respond and provide assistance, as well as to adjust traffic patterns.

The Moscow authorities are holding a tender for “the implementation of work to create a unified automated system for aggregating and processing information on incidents on the road network,” which will be called “PROSobitye.” The city is allocating 285 million rubles for this. The terms of reference for the tender say that the neural network being created will recognize emergency situations that are not formally violations of traffic rules. The documentation mentions such incidents as “vehicle intrusion into a prohibited area”, “pedestrian intrusion into a prohibited area”, “registration of a concentration of vehicles moving at an average speed significantly lower than the normal speed for a given area”, “vehicle movement at a speed significantly lower than the speed of the main flow.” Non-standard situations, explained to Kommersant at the Moscow Traffic Management Center (TCO), will also include accidents, fire, parked cars in the traffic lane, objects on the road.

“Information will be automatically transmitted instantly to emergency services and other city services for response and assistance,” explained the data center. “Thanks to the system, an analysis of the traffic situation will be carried out. For example, information about difficulties will be automatically transmitted to navigation services and to the emergency information display. We will be able to quickly respond to all incidents by automating the coordination of all services, inform drivers themselves about any changes on the road, and receive analytical and statistical reports.” PROSObytie will also record traffic violations, but exclusively for analytics and correction of traffic management patterns that lead to driver errors on the roads.

At the beginning of 2022, we recall that the data processing center signed a contract for 4.2 billion rubles. with the Spetsdorproekt company to deploy an intelligent video surveillance system capable of detecting accidents, traffic jams, animals, garbage or abandoned cargo on the roadway using a neural network. As part of the contract, 1.4 thousand video analytics cameras were installed on the Moscow Ring Road and a number of outbound highways. Now the data center is creating a citywide system that will include the platform previously created by Spetsdorproekt, and will also use data from other city cameras. We would like to remind you that the data center now has more than 3.8 thousand complexes for recording violations and more than 5.7 thousand different television cameras (including 1.4 thousand cameras installed in 2022).

Coordinator of the Blue Buckets movement Petr Shkumatov says that the system is potentially useful for the city, since it will allow identifying not only road accidents, but also pedestrians running onto the road and other similar events. At the same time, the expert believes that in the conditions of capital traffic there is a high risk of “false” alarms of the system. For example, often on city roads you can encounter a situation where the right lane is driving too slowly (vehicles are stuck in a traffic jam turning onto a bridge), and the left lanes are moving quickly: the neural network can recognize this as a “cluster of vehicles moving at an average speed, significantly lower than the normal speed for this area.”

The head of the Probok.net expert center, Alexander Shumsky, expects that PROSystem will make it possible to more effectively deal with so-called secondary traffic accidents, when, as a result of a minor accident without injuries, drivers stop to fill out documents (or while waiting for the traffic police), after which they are At full speed, other road users crash into each other. At the beginning of 2023, as Kommersant reported, the data processing center began providing insurers with video from cameras on the Moscow Ring Road and urging motorists to quickly remove their cars from the roadway, completing registration outside the road. As a result, according to Mr. Shumsky’s estimates, the number of “secondary” accidents has decreased by 20–25%, but the new video analytics system will reduce their number more radically due to a quick response to them (the Data Center sends a “Road Patrol” car to an accident).

Ivan Buranov