Diffusion guided training improves multimodal autonomous driving control
Diffusion-2BC: Hybrid Diffusion and Regression Training for Offline Behavior Cloning in Autonomous Driving
RoboticsMachine Learning
Summary
Self-driving cars can learn by copying human drivers, but sometimes one situation can have many good driving choices, which makes simple copying less effective. The authors combined two training methods: a diffusion method that can handle multiple good actions and an extra guiding signal that helps the system learn more reliably from images and control data. They tested their approach in simulated driving environments and found it improved driving accuracy and decision making compared to simpler methods. This hybrid method keeps the ability to choose from multiple routes while making the driving behavior more stable.
What this means in practice
- •For autonomous vehicle developers: Improve offline training of driving policies to better handle multiple possible driving actions from visual data with more stable control.
- •For robotics navigation teams: Enhance navigation systems by combining multimodal action prediction and regression training to increase closed-loop reliability in simulated environments.
Authors
Bruno Maciel Machado, Eric Aislan Antonelo
Abstract
Behavior cloning provides an offline route to autonomous-driving policy learning, but mean-squared-error regression is poorly matched to demonstrations in which one observation admits several valid actions. Diffusion policies can represent conditional multimodal action distributions, yet their closed-loop performance may be unstable when visual features and control are learned from limited data. This paper presents Diffusion-2BC, which combines a diffusion denoising objective with an auxiliary deterministic behavior-cloning loss over a shared visual encoder. The auxiliary branch is used only during training; inference remains diffusion-based. The proposed method is evaluated in the controlled Claw environment and in bird's-eye-view CARLA navigation, including route-conditioned driving, route-free navigation through multiple intersections, and cross-map evaluation from Town01 to Town02. In the Claw task, Diffusion-2BC reduced the mean mask-distance error by approximately 10% relative to a diffusion-based behavior-cloning baseline and by 85% relative to standard deterministic behavior cloning. In route-free CARLA, Diffusion-2BC traveled substantially farther before termination under the evaluation protocol than both baselines in Town01 and Town02. Additional qualitative rollouts revealed distinct route choices, showing the multimodal behavior of the proposed diffusion-based agent. The results indicate that an auxiliary regression signal can improve the closed-loop reliability of diffusion behavior cloning while preserving multimodal prediction in the controlled benchmark.