Designed at ETU "LETI": The world's best intelligent solutions for managing driverless vehicles
Konstantin Chaika, postgraduate student of the Department of Software Engineering and Computer Applications of ETU “LETI,” became the absolute champion of AI Driving Olympics as a member of the Russian national team.
From May 20 to 24, the International Conference on Robotics and Automation (ICRA) took place in Montreal (Canada), organized by the Institute of Electrical and Electronics Engineers (IEEE).
Within the framework of the event, on May 22, the AI Driving Olympics final took place. It is a competition on applying AI to the operation of self-driving cars, which took place for the second time. The championship was based on the Duckietown platform - a testing ground for autonomous traffic algorithms, which is a reduced model of the urban transport environment and includes road markings, vehicles, traffic lights, and buildings.
The joint team of students called JBRRussia participated in the competition, which included Konstantin Chaika, postgraduate student of the Department of Software Engineering and Computer Applications of ETU “LETI,” and students of ITMO University, Higher School of Economics Campus in St. Petersburg, and Computer Science Center. Team leaders are Kirill Krinkin, Head of the Department of Software Engineering and Computer Applications and Alexey Shpilman, Senior Lecturer at the Informatics Department of the Higher School of Economics Campus in St. Petersburg.
“Unfortunately, I could not go to Canada for the final. Despite this, participation in competitions allowed me to make sure once again that if you put in the effort and try, everything will work out. I am very grateful to my mentor, Kirill Krinkin, for the fact that he interested me in the Duckietown platform and gave me the opportunity to study it,” Konstantin Chaika, postgraduate student of the Department of Software Engineering and Computer Applications of ETU “LETI,” said.
The Olympics consisted of three disciplines. The first was autonomous movement in the lane according to the markings in compliance with the traffic rules. Here, developers had to present a solution, thanks to which the robot could ride smoothly in a lane without going to the oncoming lane or the curb and not taking off on bends. At the second stage, the task was to drive in a lane with other vehicles. It was necessary to pass the road without encountering robots traveling in the opposite lane and in the same direction. The third discipline was driving around the city with other vehicles. It involved solving the problems of passing intersections without interfering with other vehicles.
Each solution was tested in a simulator and on a real robot. When grading, the jury took into account three parameters: the duration of the work of the robot, the distance traveled, and the degree of deviation from the markings.
From February to April, students developed solutions for the competitions with the support of the association JetBrains Research according to a template provided by the organizers, tested it on the simulator, and corrected the shortcomings.
“We understood that we need to start preparing for competitions as soon as possible since we had to take into account a lot of nuances. In addition, the team had new people who needed time to figure out the subject area. By the beginning of the competition in Canada, we had a solution that worked well in the simulator and on a real robot. However, during testing on the Duckietown platform, errors were revealed that we corrected remotely: some of the guys were in Canada, the others were in Russia.”
The JBRRussia team managed to create a solution that showed good results both in the simulator and on the physical robot. The Russian team took the first place, ahead of their main competitor - the MYF team from China.
“We achieved success through a balanced combination of machine learning methods with classical computer vision-based solutions. The participation of ETU “LETI” students and graduate students in such international events as ICRA allows them to integrate into the current scientific agenda and, therefore, to solve real problems facing the industry of the future. A victory in such competitions indicates a high level of training of our students.”