Assetto Corsa Gym Large-scale Autonomous Racing Simulation Benchmark

The Assetto Corsa Gym dataset was launched in 2024 by the University of California, San Diego, Graz University of Technology, Unimore and Know-Center GmbH.A Simulation Benchmark for Autonomous Racing with Large-Scale Human Data".

The dataset collects 64 million steps of racing driving data, of which 2.3 million steps come from human drivers of different driving levels and the rest come from Soft Actor-Critic (SAC) strategies. The data collection involved 15 drivers, who completed at least five laps per track and per car. Participants included one professional e-sports driver, four experts, five amateur drivers, and five beginners. The following figure summarizes the data:

Supported tasks and leaderboards

• Autonomous driving
• Reinforcement Learning
• Behavioral cloning
• Imitation Learning