Russian Students Develop AI Program for Autonomous Transport

Students from ITMO University in St. Petersburg have developed one of the most advanced AI-based navigation solutions in Russia. According to the university’s press service, their project won first place at the prestigious Artificial Intelligence and Data Analysis Olympiad, outperforming 248 teams from 14 countries.

The team, consisting of Daria Ledeneva and Timur Ionov, created a sophisticated neural-network–based mapping model designed to optimize autonomous transport navigation. The challenge was issued by Yandex, one of the major industry players working on self-driving technologies.

To achieve high performance, the students combined several neural network models capable of processing camera imagery and detecting obstacles. The resulting system transfers this visual information onto a top-down digital map, highlighting obstacles and navigational constraints.

The students enhanced model accuracy by teaching it to interpret not only clear object boundaries, but also blurred contours — a crucial element for autonomous navigation in real-world conditions. As Ledeneva explained, their approach significantly improved obstacle recognition.

In terms of performance, the model achieved an Intersection over Union (IoU) score of 0.564 on the private leaderboard, securing first place for their team AI Capybara. This marks the second consecutive year that ITMO students have won the competition, underscoring the strong capabilities of Russian young specialists in AI and data science.