Automatic accurate stopping and safe passenger transfer based on Computer Vision and AI-enhanced techniques

CAF Signalling is involved in different research projects related to Computer Vision (CV) and Artificial Intelligence (AI) enhanced systems develop in order to reach a higher autonomy in urban vehicles and align them with railway European normative. The objective is to apply CV&AI techniques to improve different autonomous train operation functionalities as precision stop, visual odometry, rolling stock coupling operation or person and obstacle detection-identification in railroads.

However, as many companies across the sector, CAF Signalling is facing up different computational capabilities challenges for CV&AI-enhanced autonomous train operation which needs real-time & safety-critical computing platforms for correct performance.

The future of CV&AI breakthroughs in railway sector will require large arrays of memory devices at the same accuracy as a Graphical Processing Unit (GPU)-based system, hardware accelerators and new platforms. These achievements will expand the scale of CV&AI processing-calculations making them larger and faster (this means energy-efficiency must improve dramatically).

CAF Signalling will use the SELENE approach on AI-enabled computing platforms to execute some functionalities developed in CV&AI field for autonomous train operation. More precisely, SELENE’s project use case will focus on:

1. Automatic platform detection. It will detect platform area based on train localization information (odometry sensors, balise information…) and different visual pattern (visual sensors) detection/identification (characteristic patterns which identifies train platforms). Platform detection functionality will enable CV&AI based automatic train approximation to accurate train stop.
2. Automatic accurate stop at door equipped platforms aligning the vehicle and platform doors. It will perform precise localisation inside platform area using visual patterns detection, identification and tracking in order to reach accurate stopping point and managing automatic train operation (traction and brake commands, ATO functionality). The visual patters will be designed and chosen to maximize the detection and identification processes results in any possible lightness and meteorological conditions. On the other hand, these patterns will be installed according predefined precise distances to obtain physical accurate measurement from correct calibrated visual sensors.
3. Safe passenger transfer. It will manage automatic safe door enabling (ERMTS functionality) making sure the train is completely stopped in the platform area (using visual sensors) avoiding door opening operation if the train and platform doors are not precisely aligned.

As expected results, CAF Signalling will market in 3-5 years more accessible and flexible solutions aligned with social sustainability and mobility concerns fulfilling urban operators need. If urban vehicles (metro) gain autonomy system development costs are reduced (install and maintenance costs) and operation flexibility is gained. Furthermore, applying autonomous vehicle techniques based on artificial vision and artificial intelligence to railway sector would allow public administrations to invest in urban transport that would offer society the higher safety levels and at the same time increase transport capacity.

The goal of CAF for this use-case is:

• Give to vehicles autonomy and decision-making capabilities so they can observe and interpret the environment in an independent manner, complementing the information already received from railroad signaling modes.
• Reduce installing and maintenance costs by lowering complexity and Price with new optical sensing and increasing installations’ the life cycle.
• Increase flexibility in different railway operations that are attached to delimited areas and delimited time slots depending on the type of railroad and its configuration.
• Variable calculations and operation executions enhancement both in precision and speed with new optical sensors information. 
• Increase railway systems safety level at the same time that their exploitation capacity is increased by new and more precise measurements and more flexible operations.