Hierarchical Motion Planning Adaptation Via Guided Diffusion
Woosung Kim, Nicola Bezzo
AI summary
Problem
Mobile robots operating in partially known environments face unpredictable dynamic degradation and environmental shifts that invalidate pre-planned trajectories, yet existing adaptation methods rely on precise system models or costly online retraining.
Approach
A two-layer framework generates a safe global path at the high level, while a low-level conditional diffusion model refines it into a dynamically feasible trajectory by conditioning on online-estimated dynamic limits and applying guided denoising with local potential fields.
Key results
- Enables zero-shot runtime adaptation to asymmetric yaw-rate degradation without retraining
- Integrates guided denoising with local potential fields for real-time obstacle avoidance
- Validated through complex simulations and real-world hardware experiments under fault conditions
- Maintains global path alignment while ensuring dynamic feasibility under changing constraints
Why it matters
Empowers mobile robots with resilient, persistent autonomy for long-duration missions in unpredictable real-world settings where hardware degradation and environmental changes are inevitable.
Abstract
Mobile robots deployed for persistent operations in partially known environments need to be able to recover and adapt against unforeseen changes in dynamics, e.g., due to failures, or external disturbances. This paper presents a novel hierarchical framework capable of zero-shot adaptation to environmental and dynamic changes. At the high level, an abstract planner generates a collision-free global path, adapting to degraded mobility by inflating a dynamic safety buffer around obstacles to ensure the route remains navigable. At the low level, a concrete planner employs a conditional Denoising Diffusion Probabilistic Model (DDPM) to refine the abstract path into a smooth, executable trajectory. The key to our approach is conditioning the diffusion model’s generation process on the robot’s online-estimated dynamic limits. Our framework’s effectiveness and robustness are validated in both complex simulations and real-world hardware experiments, demonstrating its ability to ensure mission success under unstructured and unexpected fault situations.