Self-Imitated Diffusion Policy for Efficient and Robust Visual Navigation
For robotic visual navigation, SIDP addresses the inefficiency and sub-optimality of diffusion policies trained with imitation learning, enabling faster and more robust real-time deployment.
The paper proposes Self-Imitated Diffusion Policy (SIDP) for visual navigation, which uses reward-guided self-imitation and curriculum learning to improve trajectory quality and efficiency. SIDP achieves 2.5× faster inference (110ms vs 273ms) than NavDP on Jetson Orin Nano and outperforms prior methods in simulation and real-world tests.
Diffusion policies (DP) have demonstrated significant potential in visual navigation by capturing diverse multi-modal trajectory distributions. However, standard imitation learning (IL), which most DP methods rely on for training, often inherits sub-optimality and redundancy from expert demonstrations, thereby necessitating a computationally intensive "generate-then-filter" pipeline that relies on auxiliary selectors during inference. To address these challenges, we propose Self-Imitated Diffusion Policy (SIDP), a novel framework that learns improved planning by selectively imitating a set of trajectories sampled from itself. Specifically, SIDP introduces a reward-guided self-imitation mechanism that encourages the policy to consistently produce high-quality trajectories efficiently, rather than outputs of inconsistent quality, thereby reducing reliance on extensive sampling and post-filtering. During training, we employ a reward-driven curriculum learning paradigm to mitigate inefficient data utility, and goal-agnostic exploration for trajectory augmentation to improve planning robustness. Extensive evaluations on a comprehensive simulation benchmark show that SIDP significantly outperforms previous methods, with real-world experiments confirming its effectiveness across multiple robotic platforms. On Jetson Orin Nano, SIDP delivers a 2.5$\times$ faster inference than the baseline NavDP, i.e., 110ms VS 273ms, enabling efficient real-time deployment.