Topology-Guided ORCA: Smooth Multi-Agent Motion Planning in Constrained Environments

TL;DR

The paper tackles the problem of inefficient and sometimes infeasible motion planning for multiple agents in environments with static obstacles when using ORCA. It introduces Topology-Guided ORCA, which uses a Medial Axis Transform to create a topological graph of the traversable space and augments this graph for each agent with its start and goal to generate a waypoint sequence that guides ORCA. Agents follow the waypoints via ORCA between successive goals, enabling smoother and more natural motion around obstacles; replanning occurs as goals are reached. Across constrained crowd simulations, the method outperforms standard ORCA by achieving higher velocities, fewer frozen frames, and fewer stuck agents, demonstrating potential for training constrained social navigation policies.

Abstract

We present Topology-Guided ORCA as an alternative simulator to replace ORCA for planning smooth multi-agent motions in environments with static obstacles. Despite the impressive performance in simulating multi-agent crowd motion in free space, ORCA encounters a significant challenge in navigating the agents with the presence of static obstacles. ORCA ignores static obstacles until an agent gets too close to an obstacle, and the agent will get stuck if the obstacle intercepts an agent's path toward the goal. To address this challenge, Topology-Guided ORCA constructs a graph to represent the topology of the traversable region of the environment. We use a path planner to plan a path of waypoints that connects each agent's start and goal positions. The waypoints are used as a sequence of goals to guide ORCA. The experiments of crowd simulation in constrained environments show that our method outperforms ORCA in terms of generating smooth and natural motions of multiple agents in constrained environments, which indicates great potential of Topology-Guided ORCA for serving as an effective simulator for training constrained social navigation policies.

Figures (1)

• Figure 1: (a) The graph representation of an environment including three static obstacles and (b) An agent's augmented graph. The green circle is the agent, and red lines show its path toward the goal. As one can see, the generated path guides the agent to move around the obstacles.