Improvement of the smart tram in Pilsen continues

Cooperation Science Public

Stop in front of an obstacle, interact with the environment or react to different situations. All of these skills can be performed by the smart tram. Right now, they are testing methods for analysing safety-critical functions.

The modern tram is equipped with special units that allow it to communicate with the transport infrastructure and orient itself in space thanks to cameras and lidars (Light Detection And Ranging - a method for remote distance measurement). Using up-to-date and highly accurate map data, it can react to unforeseen events while driving. These include, for example, obstacles on the track or pedestrians crossing the road. Since July this year, the tram has been running on a test circuit in Pilsen.

Some elements of the tram were developed within the DIDYMOS research project. This project is partly followed by experts from the Department of Cybernetics at the Faculty of Applied Sciences (FAV), headed by Lukáš Picek. Together with colleagues and in cooperation with Škoda Digital and Škoda Transportation, they are working on the so-called Smart Depot project. The depot and polygon in the Finnish city of Tampere will be used to test automated and autonomous rolling stock. Sensor elements will be installed in the test section to automatically detect obstacles using lidar or cameras. "The project aims to design and test in a real environment methods for analyzing the safety-critical functions of AI algorithms for locating and detecting obstacles usable in autonomous tram control," explains Lukáš Picek.

Currently, a relatively large number of systems address the use and design of conventional and advanced computer vision and machine learning methods in autonomous vehicle control and robotics. In most cases, simpler types of sensors, such as radar or ultrasound, are used for obstacle detection. This is due to the complexity of AI algorithms and the vagueness of standards that certification of systems using AI algorithms requires.

"The safety of these methods has been little explored in the autonomous vehicle control and robotics environment. Therefore, our team focuses on the basic process definitions for developing and testing AI methods in safety-critical applications. Of course, we do not just stick to definitions, but try to validate our methods in practice in conjunction with the Škoda Group," says Lukáš Picek. From the data obtained, it will be possible to clearly define how computer vision and machine learning algorithms can be used and implemented in safety-critical functions, how to determine the parameters important for determining their reliability, and how to prove and guarantee them in live operation.

This project is funded with state support from the Technology Agency of the Czech Republic and the Ministry of Transport of the Czech Republic under the Transport 2030 Programme.

  

Gallery


Faculty of Applied Sciences

Martina Batková

23. 10. 2024