Multi-Agent Pathfinding Under Team-Connected Communication Constraint via Adaptive Path Expansion and Dynamic Leading

TL;DR

This work tackles multi-agent pathfinding under a strict team-connected communication constraint, requiring a spanning communication tree to persist throughout motion. It introduces a two-level planner, APEDL, combining Adaptive Path Expansion (multi-stage path growth with collision-at-goal trimming) and Dynamic Leading (leader reassignment during expansion) within a Team Communication Tree to manage planning progress. Empirical results show the planner scales to up to 25 agents under limited-range communication and maintains high success rates across five environment types, outperforming baselines that fail in dense, constrained scenarios. The approach provides a practically efficient, robust solution for communication-constrained MAPF, with planned extensions toward completeness and continuous kinodynamic planning.

Abstract

This paper proposes a novel planning framework to handle a multi-agent pathfinding problem under team-connected communication constraint, where all agents must have a connected communication channel to the rest of the team during their entire movements. Standard multi-agent path finding approaches (e.g., priority-based search) have potential in this domain but fail when neighboring configurations at start and goal differ. Their single-expansion approach -- computing each agent's path from the start to the goal in just a single expansion -- cannot reliably handle planning under communication constraints for agents as their neighbors change during navigating. Similarly, leader-follower approaches (e.g., platooning) are effective at maintaining team communication, but fixing the leader at the outset of planning can cause planning to become stuck in dense-clutter environments, limiting their practical utility. To overcome this limitation, we propose a novel two-level multi-agent pathfinding framework that integrates two techniques: adaptive path expansion to expand agent paths to their goals in multiple stages; and dynamic leading technique that enables the reselection of the leading agent during each agent path expansion whenever progress cannot be made. Simulation experiments show the efficiency of our planners, which can handle up to 25 agents across five environment types under a limited communication range constraint and up to 11-12 agents on three environment types under line-of-sight communication constraint, exceeding 90% success-rate where baselines routinely fail.

Figures (9)

• Figure 1: Fixed leader and single-expansion of agent paths fail in scenarios a) & b). Our proposed techniques: adaptive path expansion and dynamic leading handle both (c). If $a_1$'s leading causes the team to get stuck (a), dynamic leading allows another agent take over as the new leader. When the leader and followers move to different directions (b), adaptive path expansion allows the stuck-agent to resume planning in next expansion.
• Figure 2: An illustration of one-step 's expansion with 3 agents. Agents $a_1$, $a_2$, and $a_3$, start at $s_1$, $s_2$, and $s_3$, and target their goals $g_1$, $g_2$, and $g_3$ while avoiding the obstacles (black) and other agents. From node $v_1$ in $\mathcal{T}$$_{tc}$ (step 1a), the planner expands trajectories for each agent initially with a planning order. Our first technique, adaptive path expansion---from step 1b to 5, grows and refines agent paths. The second technique, dynamic leading, at step 4 and 5, allows $a_3$ and $a_2$ to reach their goals. The tree $\mathcal{T}$$_{tc}$ is expanded at step 6.
• Figure 3: Out-of-communication-at-goal situation (a) and how ${{{\textsc{IsCommAtGoal}}}}$() works (b). In both situations, the planning order is ($a_3, a_2, a_1$) and the leader is changed at $t_{i+2}$.
• Figure 4: Success-rate, Runtime, and Travel Distances of our planners: and -e with , , , , and on 5 environment types under constraint. The comm. range is 15 m and max-runtime is 5 s. The runtime and travel distances are shown by means and standard deviations.
• Figure 5: Performance of as varying the number of path expansions in Maze environments (Env.5). One-time path expansion ($m=1$) fails as the number of agents increase. Adaptive path Expansion (ape) ($m>1$) can compute paths efficiently even with Goal Configuration in Long-Thin Shape.