D4orm: Multi-Robot Trajectories with Dynamics-aware Diffusion Denoised Deformations

2025-03-15

AI Summary

To address the challenge of simultaneously ensuring dynamical feasibility, collision-free navigation, and real-time performance in multi-robot trajectory planning, this paper proposes a training-free diffusion denoising optimization framework. The method integrates the progressive denoising mechanism of diffusion models into trajectory optimization, where the score function is constructed via dynamics-forward simulation and Monte Carlo gradient approximation, and refined through GPU-accelerated iterative deformation optimization. Key contributions include: (i) the first application of diffusion-based generative modeling to kinematic/dynamic-constrained trajectory generation; (ii) real-time refinement with arbitrary termination time; (iii) cross-platform dynamical generalizability; and (iv) zero-shot deployment on physical systems. Experiments demonstrate that the method achieves ~3× speedup over MPPI in planning 16-robot formations in 2D/3D environments. Furthermore, it enables zero-shot deployment on eight quadrotors, validating both practical efficacy and generalization capability.

Technology Category

• Intelligent Robots: Motion and Path Planning
• Search and Optimization: Sampling/Simulation-based Search
• Planning, Routing, and Scheduling: Replanning and Plan Repair

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

This work presents an optimization method for generating kinodynamically feasible and collision-free multi-robot trajectories that exploits an incremental denoising scheme in diffusion models. Our key insight is that high-quality trajectories can be discovered merely by denoising noisy trajectories sampled from a distribution. This approach has no learning component, relying instead on only two ingredients: a dynamical model of the robots to obtain feasible trajectories via rollout, and a score function to guide denoising with Monte Carlo gradient approximation. The proposed framework iteratively optimizes the deformation from the previous round with this denoising process, allows extit{anytime} refinement as time permits, supports different dynamics, and benefits from GPU acceleration. Our evaluations for differential-drive and holonomic teams with up to 16 robots in 2D and 3D worlds show its ability to discover high-quality solutions faster than other black-box optimization methods such as MPPI, approximately three times faster in a 3D holonomic case with 16 robots. As evidence for feasibility, we demonstrate zero-shot deployment of the planned trajectories on eight multirotors.

Problem

• Generates collision-free multi-robot trajectories
• Uses diffusion models for trajectory denoising
• Supports dynamic models and GPU acceleration

Innovation

• Uses diffusion models for trajectory denoising
• Relies on dynamical models and score functions
• Supports GPU acceleration for faster computation