Rapid and Safe Trajectory Planning over Diverse Scenes through Diffusion Composition

TL;DR

This work tackles real-time, safe trajectory planning in complex, diverse scenes by learning from MPC-generated demonstrations and deploying diffusion-based planners. It introduces a diffusion composition strategy to generalize to unseen environments without retraining, coupled with a lightweight rule-based safety filter to guarantee collision avoidance. Offline MPC-based data collection yields kinematically feasible demonstrations, while online diffusion synthesis maps raw sensor inputs to executable trajectories with high stability. Real-world validation on an F1TENTH platform demonstrates practical viability, rapid planning (approx. 0.21 s), and robust safety through model composition and the safety filter, enabling reliable operation in dynamic scenarios.

Abstract

Safe trajectory planning in complex environments must balance stringent collision avoidance with real-time efficiency, which is a long-standing challenge in robotics. In this work, we present a diffusion-based trajectory planning framework that is both rapid and safe. First, we introduce a scene-agnostic, MPC-based data generation pipeline that efficiently produces large volumes of kinematically feasible trajectories. Building on this dataset, our integrated diffusion planner maps raw onboard sensor inputs directly to kinematically feasible trajectories, enabling efficient inference while maintaining strong collision avoidance. To generalize to diverse, previously unseen scenarios, we compose diffusion models at test time, enabling safe behavior without additional training. We further propose a lightweight, rule-based safety filter that, from the candidate set, selects the trajectory meeting safety and kinematic-feasibility requirements. Across seen and unseen settings, the proposed method delivers real-time-capable inference with high safety and stability. Experiments on an F1TENTH vehicle demonstrate practicality on real hardware. Project page: https://rstp-comp-diffuser.github.io/.

Figures (6)

• Figure 1: Diffusion Composition Enables Efficient, Safe Planning with Practical Real-world Performance. An individual diffusion model cannot ensure safe trajectory planning in out-of-distribution scenarios, whereas composing multiple diffusion models can achieve safety during generalization. Dashed boxes indicate obstacles that do not exist during training. Validation on the F1TENTH platform shows that trajectories planned by the composed diffusion model offer excellent safety while maintaining computational efficiency, demonstrating effectiveness for practical real-world applications.
• Figure 2: Diffusion Composition. Trajectory planning using the energy model as a surrogate model for the potential field.
• Figure 3: The Overall Framework of the Proposed Rapid and Safe Trajectory Planning (RSTP) Method.Offline Dataset Generation (Left): the ItCA liRapidIterativeTrajectory and MPC-based methods provide kinematically feasible trajectory datasets for training. Diffusion Composition (Middle): individual diffusion models can be flexibly composed to tackle novel scenarios not covered in the training data. Online Inference and Control (Right): The ego vehicle performs real-time inference from its current pose, and the safety filter selects the optimal trajectory per cycle for tracking.
• Figure 4: Diffusion Model Can Ensure Kinematic Constraints. The RSTP method can plan safe trajectories involving large curvature \ref{['fig:static']} and gear shifting points (GSP) \ref{['fig:static2']}.
• Figure 5: The Composed Model Flexibly Avoids Obstacles in Unseen Scenarios by Adjusting Velocity. Comparison of F1TENTH's velocity under different diffusion models in $CS_1$. The composed model is able to decelerate to avoid static obstacles, whereas the dynamic model (DyM1) fails to avoid the additional static obstacles.