Hierarchical Planning for Heterogeneous Multi-Robot Routing Problems via Learned Subteam Performance

Abstract

This paper considers a particular class of multirobot task allocation problems, where tasks correspond to heterogeneous multi-robot routing problems defined on different areas of a given environment. We present a hierarchical planner that breaks down the complexity of this problem into two subproblems: the high-level problem of allocating robots to routing tasks, and the low-level problem of computing the actual routing paths for each subteam. The planner uses a Graph Neural Network (GNN) as a heuristic to estimate subteam performance for specific coalitions on specific routing tasks. It then iteratively refines the estimates to the real subteam performances as solutions of the low-level problems become available. On a testbed problem of a heterogeneous multi-robot area inspection problem as the base routing task, we empirically show that our hierarchical planner is able to compute optimal or near-optimal (within 7%) solutions approximately 16 times faster (on average) than an optimal baseline that computes plans for all the possible allocations in advance to obtain precise routing times. Furthermore, we show that a GNN-based estimator can provide an excellent trade-off between solution quality and computation time compared to other baseline (nonlearned) estimators.

A Concurrent Mission-Planning Methodology for Robotic Swarms Using Collaborative Motion-Control Strategies

“…Existing approaches to SRS mission-planning can be categorized based on the operational limitations considered in their formulation, including computational, locomotion, and localization ones. Studies that are most similar to the work proposed in this paper, typically, address locomotion [10][11][12][13][14][15][16] and localization [17][18][19][20] limitations of the swarm. These along with other works related to swarm missionplanning are reviewed below in detail.…”

“…The locomotion limitations of swarm robots would affect the executability of the trajectories that are planned for a task-allocation solution considered. In such cases, the literature advocates that if a robot trajectory is not executable within the limits of its capabilities, changes must be made to the mission plan at the task-allocation level [10][11][12][13][14][15][16]. This represents a constrained mission-planning problem and calls for a strategy that searches for the optimal task-allocation and swarm trajectories concurrently.…”

On the role and opportunities in teamwork design for advanced multi-robot search systems

“…At last we discuss the topic of learning sub-team performance as an additional future research direction for advancing multi-robot searching systems. In Banfi et al. (2022) , hierarchical planning for heterogeneous multi-robot routing problems via learned sub-team performance is examined in a different context than search and rescue applications however ideas drawn from the paper may be applied in this context as well.…”

A Sampling-Based Approach for Heterogeneous Coalition Scheduling with Temporal Uncertainty

“…In our setting, we assume that robots belonging to a heterogeneous team have Fig. 1: Polypixel, a simulated urban environment used for our experiments (as seen in [7]), with its graph-based abstraction superimposed. Vertices show locations where a task has been assigned to one or more robots of a heterogeneous team.…”

“…We show that the use of CS-HSSRG for dealing with the scheduling part of the problem within the stateof-the-art multi-agent coordination planner GRSTAPS [12] results in decisions about what tasks to include in the plan and what robots to allocate to those tasks that are robust to temporal uncertainty. The proposed approach is fundamentally centralized and offline, similarly to many recent works in the same area [13,3,14,12,7], and centralization of the planner is often imposed by operational constraints. However, our approach is experimentally efficient enough to allow online replanning phases which may be triggered by different events such as subteams becoming disconnected.…”