NavCMPO: Critic-Guided MeanFlow Policy Optimization for Adaptive Navigation

2026-07-16

AI Summary

This work addresses the high inference latency of end-to-end diffusion policies and the reliance of behavior cloning on expert demonstration quality in vision-based navigation without maps. To overcome these limitations, the authors propose a two-stage adaptive navigation framework: first generating coarse trajectories efficiently using few-step MeanFlow, then refining them through critic-guided trajectory optimization and PPO-based reinforcement learning to enhance obstacle avoidance. Additionally, an obstacle proximity prediction auxiliary task is introduced to strengthen spatial awareness in visual representations. Evaluated on the InternVLA-N1 benchmark, the method achieves a 74.7% average success rate—6.4 percentage points higher than NavDP—with inference latency reduced from 85 ms to 60 ms. Successful sim-to-real transfer is further demonstrated on the Unitree Go2 robotic platform.

Abstract

End-to-end diffusion-based policies have demonstrated strong performance in mapless visual navigation, but their iterative denoising process introduces substantial inference latency, while behavior cloning limits performance to the quality of expert demonstrations. We present NavCMPO, a two-stage adaptive navigation framework that combines few-step MeanFlow trajectory generation, critic-guided refinement, and reinforcement learning fine-tuning. During pre-training, an obstacle proximity prediction task encourages the visual representation to capture obstacle-aware spatial information. To compensate for the degradation in obstacle avoidance caused by few-step generation, Critic-Guided Trajectory Refinement (CGTR) uses gradients from a critic trained with obstacle-point-cloud supervision to refine intermediate trajectories. During adaptation, the MeanFlow policy is fine-tuned using Proximal Policy Optimization with behavior-cloning regularization, while the critic is updated to accommodate embodiment-specific observation changes. Under a matched training budget on the InternVLA-N1 benchmark, NavCMPO achieves an average success rate of 74.7\%, exceeding the retrained NavDP baseline by 6.4 percentage points, while reducing inference latency from 85\,ms to 60\,ms. Experiments on a Unitree Go2 further demonstrate effective sim-to-real transfer.

Problem

• visual navigation
• diffusion policy
• inference latency
• behavior cloning
• obstacle avoidance

Innovation

• MeanFlow
• Critic-Guided Trajectory Refinement
• Diffusion Policy
• Obstacle-Aware Representation
• Sim-to-Real Transfer