DriveMLP shows simple training beats complex models in autonomous driving
DriveMLP: When Training Strategy Outperforms Model Complexity in Autonomous Driving
Machine LearningRobotics
Summary
Motion planning in self-driving cars is hard because small mistakes can pile up when the car makes decisions on its own. The authors show that how you train the model can matter more than making the model bigger or more complex. They created a simple method called Expert-Guided Reinforcement Learning (EGRL) that uses expert driving data to guide the training. Using this, their lightweight DriveMLP model performs better and runs faster than previous imitation learning methods.
What this means in practice
- •For autonomous vehicle developers: Improve driving policy training efficiency and reliability with simpler models trained by expert-guided reinforcement learning strategies.
- •For robotics software engineers: Use lightweight MLP models with expert-guided training to speed up real-time decision making in robots requiring motion planning.
Authors
Alessandro Canevaro, Julian Schmidt, Hang Yu, Ye Jin, Sajad Marvi, Georg Martius, Julian Jordan
Abstract
Learning motion planning policies that remain reliable under closed-loop execution is a central challenge in Autonomous Driving (AD). Imitation Learning (IL) scales well to large AD datasets but suffers from distribution shift, causal confusion, and compounding errors under closed-loop execution. In contrast, Reinforcement Learning (RL) typically requires handcrafted rule-based rewards and massive simulation. We study a simple closed-loop IL framework that uses RL purely as an optimization tool, guided by a trajectory-distance reward derived from logged expert data. We refer to this training strategy as Expert-Guided RL (EGRL). Through a controlled empirical study we find evidence that the training procedure, rather than model size and complexity, accounts for a substantial part of the performance gains. We then introduce DriveMLP, a lightweight MLP policy trained with EGRL, as a simple instantiation of these findings. It achieves a score of 86.51 on the nuPlan Val14 Reactive benchmark, a +3.2 point improvement over the previous best IL planner (Flow Planner), with >15x faster inference. Code is available at: https://github.com/alessandro-canevaro/DriveMLP .