Nuro researchers publish CIMRL safe autonomous-driving method
Nuro researchers released CIMRL, a method combining imitation and reinforcement learning for safety-focused autonomous-driving policy training.
Evidence notes
- CIMRL combines imitation and reinforcement learning to train autonomous-driving policies with safety constraints.
- The evaluation covers simulated and real-world driving benchmarks.
- The work was released as arXiv preprint 2406.08878 on 13 June 2024.
Company context
Nuro develops autonomous-driving technology for commercial vehicle partners, with history in zero-occupant delivery and newer robotaxi programs.