LeRobot releases L2D: 90TB Open-Source Self-Driving Dataset

AI Impact Summary

LeRobot’s Learning to Drive (L2D) dataset represents a significant advancement in self-driving AI development, offering 90+ terabytes of multimodal data collected from 30 German cities across 60 EVs. This expansive dataset, featuring diverse driving policies (expert and student) and comprehensive sensor data (RGB cameras, GPS, IMU, CAN), is designed to train end-to-end models for spatial intelligence, addressing a critical gap in open-source robotics AI resources. The data’s focus on EU driving tasks and inclusion of natural language instructions provides a robust foundation for training models capable of handling complex real-world driving scenarios.