Russia: The company, together with Russian Railways (RZD), has completed testing of computer vision systems for shunting locomotives. At the same time, it received the necessary permits (declaration of conformity with the requirements of TR CU 001/2011 and declaration of compliance with FSTEC information security requirements), which opened up the possibility of serial deliveries of systems to the EAEU countries. The equipment is planned to be produced at a factory in Tomsk, where $5 mln USD has been invested in.
The system was tested as part of a joint project Cognitive Pilot (a subsidiary of Sberbank and Cognitive Technologies) with RZD on two sections of the network. The hardware of the Cognitive Rail Pilot system was checked and tested in various parameters: electromagnetic compatibility, noise immunity, vibration resistance, compliance with climatic standards, resistance to moisture. During testing, ChME3 shunting diesel locomotives equipped with a computer vision system were operated for more than 300 hours in the test run mode with the participation of the developer’s specialists and more than 5,000 hours in controlled operation without their presence.
Olga Uskova, CEO of Cognitive Pilot, believes that this system will be in demand on the Russian market. The company expects that a tender will be announced this year for the purchase of such systems, which will involve equipping up to 500 locomotives with computer vision in 2022.
Elements of the Cognitive Rail Pilot system: 1 – a block of video cameras with an electronic climate control and maintenance system, and a three-axis vibration damping system, 2 – a specialised high-resolution millimeter-wave radar, 3 – a high-performance industrial computing block. Source: Cognitive Technologies
According to the manufacturer, the Cognitive Rail Pilot system can operate in any weather conditions (rain, snow, fog or at night) at temperatures ranging from -40 to +50 ˚C. Cameras integrated into the system can recognise objects at a distance of up to 300 m, but the company plans to increase this parameter by at least 2 times.
As to Cognitive Pilot calculations, the implementation of their system should reduce the number of incidents during shunting operations by up to 70%. In an interview with The Railway Equipment Journal, Olga Uskova stated that the system implemented a solution for training neural networks, which allows semi-automatically marking up and selecting classes of objects when preparing data for further development of AI. She also noted that Cognitive Rail Pilot does not require to do the reengineering of the external environment and to create additional infrastructure.
Screenshot from the Cognitive Rail Pilot vision cameras on the ChME3 shunting diesel locomotive. Source: Cognitive Technologies
Cognitive Pilot is implementing another rolling stock automation project with the tram-building PC Transport Systems. In 2019, the Cognitive Tram Pilot control system, which acts as an assistant to the driver in dangerous situations, was tested on the 71-931M Vityaz-M tram in Moscow. Olga Uskova said that this system is ready for serial installation with maximum speed control and collision avoidance functions.
Recently, PC TS won a tender for the supply of 57 low-floor trams to St. Petersburg. According to the terms of the competition, the rolling stock must be equipped with an automated control system, which, among other things, must provide the possibility of driver intervention if necessary. Cognitive Pilot is also working on driverless trams in China.
Testing the Cognitive Pilot vision system on the PC TS tram. Source: Cognitive Technologies
The global market for rolling stock automation technologies is developing dynamically. According to Visiongain analysts, today its volume is more than $7 bln USD and by 2031 it will exceed $21 bln. At the end of last year, Cognitive Pilot entered the shortlist of promising start-ups in the field of railway technology.
In Russia, the regulatory framework for launching driverless traffic on rail transport is already being worked out at the state level. Thus, significant attention is paid to driverless technologies in the draft new rules for the technical operation of railways, an updated version of which was published by The Ministry of Transport of the Russian Federation for public discussion in January this year. The document includes requirements for the operation of automatic and remote control systems for rolling stock, as well as the responsibility of the such systems developers and manufacturers for its work.
The Russian government also approved a strategy for the digital transformation of the country’s transport industry until 2030, in which one of the key initiatives is the launch of driverless vehicles. Already in June this year, the Ministry of Transport of Russia should prepare a draft of legislation act on highly automated and fully automated vehicles.
In terms of technology, a special modification of the ES2G-136 Lastochka EMU with a GoA3 automation level and the NIIAS computer vision system is to be certified this year. It claims the ability to recognise objects at a distance of up to 600 m. Also, the Moscow Department of Transport announced joint tests of a driverless tram with Yandex, which should take place this year at the depot, and then on city tracks at night. Toward the end of the year, it is planned to start test trips of an autonomous tram together with passengers. It should be noted that Yandex has not yet made any public statements about the development of computer vision systems for rolling stock.