Warm-Starting Optimization-Based Motion Planning for Robotic Manipulators via Point Cloud-Conditioned Flow Matching

2025-10-03

AI Summary

Real-time robot motion planning in dynamic human-robot collaboration faces dual bottlenecks: sampling-based methods suffer from poor scalability in high-dimensional spaces and require post-hoc smoothing, while optimization-based approaches are highly sensitive to initializations and prone to local minima. Method: We propose a generative trajectory initialization method based on single-view point-cloud-conditioned flow matching. Without requiring prior environmental knowledge, it directly learns an approximate optimal trajectory distribution from raw depth-camera point clouds and outputs a continuous probabilistic path as the optimization initial guess. A jointly trained point-cloud encoder and flow matching model enables end-to-end learning of trajectory priors. Contribution/Results: The method significantly accelerates convergence and improves optimization success rates, with strong cross-scenario generalization. In complex dynamic UR5e simulations, it achieves high planning success as a standalone module, reducing iteration counts substantially versus traditional and learning-based baselines—enabling safe, rapid, and efficient real-time replanning.

Technology Category

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

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Economics, Online Markets and Human Computation: Economic ramifications for generative AI infrastructure and applications

Abstract

Rapid robot motion generation is critical in Human-Robot Collaboration (HRC) systems, as robots need to respond to dynamic environments in real time by continuously observing their surroundings and replanning their motions to ensure both safe interactions and efficient task execution. Current sampling-based motion planners face challenges in scaling to high-dimensional configuration spaces and often require post-processing to interpolate and smooth the generated paths, resulting in time inefficiency in complex environments. Optimization-based planners, on the other hand, can incorporate multiple constraints and generate smooth trajectories directly, making them potentially more time-efficient. However, optimization-based planners are sensitive to initialization and may get stuck in local minima. In this work, we present a novel learning-based method that utilizes a Flow Matching model conditioned on a single-view point cloud to learn near-optimal solutions for optimization initialization. Our method does not require prior knowledge of the environment, such as obstacle locations and geometries, and can generate feasible trajectories directly from single-view depth camera input. Simulation studies on a UR5e robotic manipulator in cluttered workspaces demonstrate that the proposed generative initializer achieves a high success rate on its own, significantly improves the success rate of trajectory optimization compared with traditional and learning-based benchmark initializers, requires fewer optimization iterations, and exhibits strong generalization to unseen environments.

Problem

• Generating real-time robot motions in dynamic human-robot collaboration environments
• Overcoming local minima sensitivity in optimization-based motion planners
• Eliminating need for prior environmental knowledge in trajectory initialization

Innovation

• Flow Matching model for optimization initialization
• Single-view point cloud conditions trajectory generation
• No prior environment knowledge required for planning