Scalable Multi-Robot Motion Planning Using Workspace Guidance-Informed Hypergraphs

TL;DR

This work tackles scalable multi-robot motion planning in congested environments with narrow passages by extending the Decomposable State Space Hypergraph (DaSH) framework into Workspace Guided-DaSH. The method injects workspace skeleton guidance to structure the high-level task space and translates it into a low-level motion hypergraph, enabling coordinated planning of robot groups only when necessary and supporting kinodynamic feasibility via Kinodynamic RRT. Empirical results show that WG-DaSH scales to up to $128$ robots in challenging environments, outperforming several baselines in planning time and success rate, while maintaining solution quality via post-processing trajectory optimization. The approach offers a practical, topology-informed hybrid planning paradigm that combines global coordination with localized, guided search to address complex, narrow-passage MRMP problems with large robot teams.

Abstract

In this work, we propose a method for multiple mobile robot motion planning that efficiently plans for robot teams up to 128 robots (an order of magnitude larger than existing state-of-the-art methods) in congested settings with narrow passages in the environment. We achieve this improvement in scalability by extending the state-of-the-art Decomposable State Space Hypergraph (DaSH) multi-robot planning framework to support mobile robot motion planning in congested environments. This is a problem that DaSH cannot be directly applied to because it lacks a highly structured, easily discretizable task space and features kinodynamic constraints. We accomplish this by exploiting knowledge about the workspace topology to limit exploration of the planning space and through modifying DaSH's conflict resolution scheme. This guidance captures when coordination between robots is necessary, allowing us to decompose the intractably large multi-robot search space while limiting risk of inter-robot conflicts by composing relevant robot groups together while planning.

Figures (3)

• Figure 1: A comparison of the DaSH framework and Workspace Guided-DaSH showing an example task space hypergraph for each. Hypergraph vertices show which robots and/objects are present in that task space element. (left) DaSH's exhaustive construction of the task space hypergraph for a multi-manipulator rearrangement with two robots and one object. (right) Our method uses workspace guidance to induce structure onto problems that lack this natural structure, which makes construction of a minimal portion of the task space hypergraph possible. Here, robots are composed into the same planning space based on their movement along a topological skeleton (details in Sec. \ref{['sec:method']}).
• Figure 2: (a) A segment of the robot paths over the skeleton (gray) and (b) the corresponding portion of the task space hypergraph. Each task space element (circle) shows the corresponding set of robots and the relevant portion of the workspace skeleton. In (b), arrows indicate robots moving along a skeleton edge and dots indicate a decoupled intermediate state. The green and red robots are initially moving in opposite directions along the same skeleton edge. Then, red and blue reach the skeleton vertex together and go on to different edges. The yellow robot begins moving in the opposite direction of the red robot when the red robot reaches the next skeleton edge.
• Figure 3: (a) In the Warehouse environment, robots located on the top and bottom of each narrow aisle must swap places with the robot they are vertically aligned with. (b) In the Tunnels environment, robots must find and traverse multiple tunnels while moving from start positions in the back tunnel to goal positions in the front tunnel. (c) In the mining environment, robots must swap places with another robot located in an adjacent mine shaft, robot start positions are located at the front of the tunnels and at intersections between tunnels. (d-e) In the GridMaze environment, robots must move between randomly located starts (red) and goals (blue) within the maze, some of which overlap.