cs.ROOct 21, 2026

Integrating Flow Matching and Model Predictive Control for Safe Robot Navigation

Authors: Philipp Holzmann, Maik Pfefferkorn, Joao Carvalho, Ali Younes, An Thai Le, Georgia Chalvatzaki, Jan Peters, Rolf Findeisen

Organizations: Control and Cyber-Physical Systems Laboratory, Technical University of Darmstadt, Germany · Intelligent Autonomous Systems Laboratory, Technical University of Darmstadt, Germany · German Research Center for AI (DFKI), Research Department SAIROL · Hessian Centre for Artificial Intelligence · 

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

Learning-based path planners based on diffusion and flow matching can generate diverse trajectories from demonstrations but typically lack mechanisms for constraint satisfaction and collision avoidance during deployment. This work proposes a framework that combines flow-matching-based path planning with model predictive path-following control to integrate data-driven path generation with real-time, constraint-aware execution. The flow-matching model captures multimodal path distributions, while the predictive controller adapts motion online to satisfy state and input constraints and avoid obstacles using updated local sensor information. An event-triggered re-planning scheme incorporates predictive control solutions into subsequent path generation, promoting collision-free navigation in environments with previously unseen obstacles. Simulation results for mobile robot navigation demonstrate robust performance across cluttered environments. The framework provides a systematic approach to combining flow matching with predictive control for adaptive, data-driven navigation.

Figures & tables

Fig. 1: Global flow-matching-based path planning with local adaptation via model predictive path-following control (MPPFC). Given the current and goal system states, a batch of path parametrizations is sampled from the generative model. A collision-free path, considering known obstacles (gray circles), is selected (in blue), and multiple MPPFC instances with varying prediction horizons optimize the system inputs in parallel to follow it. If unforeseen obstacles (light red circle) make the current path or optimal control problem infeasible, the MPC with the shortest feasible prediction horizon is chosen and applied to the system. Once a specific condition is met, replanning is triggered to determine a new global plan biased by the latest state prediction.
Fig. 2: Path planning using a flow matching model guided by an additive gradient. Blue indicates obstacles present during data collection, whereas new obstacles are depicted in light red, with the red circle denoting a clearance margin for the mobile robot’s footprint. Generated path samples are plotted as green lines. The subfigures demonstrate: (a) unguided sampling, (b) guidance with a low scale ( γg=0.2\gamma_{\mathrm{g}}=0.2 ), and (c) guidance with a high scale ( γg=0.6\gamma_{\mathrm{g}}=0.6 ).
Fig. 3: Simulation environments. Obstacles in blue are present during the demonstrations. During deployment, unknown objects in red might appear.
Fig. 4: Simulation of a mobile robot navigating a cluttered environment. The robot (gray circle ) moves from a start (green circle ) to a goal position (red dot ). Known static obstacles are blue ( ), while previously unknown obstacles are red circles ( ) with a red boundary indicating the robot’s size margin ( ). The generative planner produces candidate trajectories (solid light-green lines ), from which a single path (dashed black line ) is selected. The executed closed-loop path is dark green ( ), and the current predicted trajectory is dashed light blue ( ). A yellow line ( ) denotes a newly chosen path from intermediate samples, while the dashed magenta line ( ) traces the intermediate path currently being followed (the prior yellow path). The dashed gray line ( ) defines the margin condition for sampling new paths.
Fig. 5: Success rates by obstacle configuration and re-planning strategy.
Fig. 6: Number of re-planning events across obstacle configurations and re-planning strategies.
More in this topic

Robot Motion Planning

Robot motion planning finds safe, feasible paths and trajectories for robots moving through constrained, changing environments.

See all 271 papers

Explore similar work

CardsList

  1. CoCoNav: Conformal Control for Safe Robot Navigation in Crowds

    Aug 7, 2026Cheng Guo, Mingzhe Ni, Zheng Liang +5Constrained Motion PlanningConformal Prediction

    Department of Computer Science, The University of Manchester, Manchester, UK. · Human-Robot Interfaces and Interaction Laboratory, Italian Institute of Technology, Genoa, Italy. · Genisom AI, Shanghai, China. +3

  2. PIER-Flow: Physics-Informed Efficient Rectified Flow for Real-Time Mobile Robot Navigation

    Jul 11, 2026Shibo Li, Zhongcheng Wang, Jiahe Cao +2Constrained Motion PlanningRobot Policy Learning

    School of Automation, Northwestern Polytechnical University, Xi’an, China · Department of Robotics, Mohamed bin Zayed University of Artificial Intelligence, Abu Dhabi SE45 05, United Arab Emirates

  3. UniConFlow: A Unified Constrained Flow-Matching Framework for Certified Motion Planning

    Jun 3, 2025Zewen Yang, Xiaobing Dai, Dian Yu +4Constrained Motion PlanningRobot Motion Generation

    Chair of Robotics and Systems Intelligence (RSI), Munich Institute of Robotics and Machine Intelligence (MIRMI), Technical University of Munich (TUM), 80992 Munich, Germany · Chair of Information-oriented Control (ITR), School of Computation, Information and Technology (CIT), Technical University of Munich (TUM), 80333 Munich, Germany · Chair of AI Planning in Dynamic Environments, Munich Institute of Robotics and Machine Intelligence (MIRMI), Technical University of Munich (TUM), 80992 Munich, Germany +2