NIIAS, the Russian Railways research institute, has presented computer vision units intended for series fitment on highly-automated Lastochka trains. The equipment was shown on the institute’s stand at the PRO//Motion. Transport Control Systems conference and exhibition. It is designed for trains operating at the third level of automation, GoA3, the last stage before fully driverless operation.
The sensor units on display include a block with four video cameras, two lidars capable of recognising infrastructure objects and detecting obstacles at a range of up to 450-500 m, and a thermal-imaging camera with a range of up to 1.5 km.
Machine vision unit developed by NIIAS for series production fitment on highly automated Lastochka trains. Source: Sergey Belov/ROLLINGSTOCK
“Over the next three years, our business plan includes a block of work to re-equip Lastochka EMUs,” said Pavel Mylnikov, deputy general director and director of NIIAS’s St Petersburg branch, showing renders of the system’s installation at the conference. “Together with Ural Locomotives [train manufacturer, part of Sinara – Transport Machines — Ed.] and other companies within the holding, we have already completed the technical design and engineering documentation for the target GoA3 train.”
According to Mylnikov, the project envisages fitting 10 such systems to trains running on the Moscow Central Circle (MCC) by the end of the year. RZD, however, told RIA Novosti that wider rollout of the technology depends on the holding’s investment programme and the matter is still under consideration.
UPD 01/09/2026: RZD confirmed 10 Lastochka EMUs to get computer vision for highly automated running on the MCC. VSM-Service, which services Lastochka and Sapsan trains in RZD’s fleet, will carry out the installation work. RZD also said testing of onboard systems and computer vision for the highest automation level, GoA4, is now in its final stage. The work is being carried out by NIIAS, part of the RZD group.
Mylnikov also said that controlled passenger operation of the GoA3 Lastochka EMU ES2G-113, which has a driver in the cab, was completed on the MCC at the end of May. The train runs independently under AI control, with the driver on board to monitor operation, open and close the doors, and take over from the system in an emergency. Over the trial period the train covered almost 200,000 km on more than 3,600 trips. “We have improved reliability 20-fold. We are now recording no more than one false trigger per 200 km, and drivers barely need to intervene in the system’s operation,” Mylnikov was quoted as saying by Gudok.
Render of the machine vision unit placement scheme on the highly automated Lastochka EMU’s carbody. Source: Pavel Mylnikov/NIIAS
At the start of 2026, RZD also outlined longer-term plans to fit 60 EMUs on the MCC with GoA4 automation, the highest, fully driverless level. Combined with the necessary infrastructure upgrades, this is expected to enable fully driverless operation from 2029. The project is driven by the goal of cutting headways on the MCC to three minutes, achievable only through greater automation.
