LeRobot goes to driving school: World’s largest open-source self-driving dataset

The Learning to Drive (L2D) dataset, developed by Yaak and LeRobot teams, is the world's largest open-source multimodal dataset aimed at creating spatial intelligence for the automotive domain. The dataset includes over a million episodes collected from 60 electric vehicles across 30 German cities for three years, featuring both expert and student driving policies. This initiative supports the training of AI models using imitation learning and reinforcement learning methods, providing comprehensive real-world driving scenarios and instructions.