What is it about?
This paper addresses a variant of the Multi-Agent Pickup and Delivery (MAPD) problem, in which a team of agents must transport items between pickup and delivery locations while avoiding collisions. MAPD arises in a wide range of applications, including automated warehouse logistics, hospital medication delivery, and construction material transport. We consider a heterogeneous MAPD setting in which agents have different capabilities and are restricted to operating within specific zones. For example, in automated logistics, certain robots may only be able to transport specific types of items or operate in designated areas; in hospital environments, some robots may be restricted to sterile zones; while in construction settings, specialized robots may be required to handle heavy materials such as steel columns. Since tasks require delivering multiple items, possibly of different types and picked up from different locations, to a single destination, they cannot usually be completed by a single agent and require cooperation among multiple agents. To address these challenges, we propose a hierarchical framework that plans item transfers across zones, coordinates handovers between agents, and calculates collision-free paths within each zone. Simulation results across several scenarios demonstrate the effectiveness of the proposed approach.
Featured Image
Photo by Homa Appliances on Unsplash
Why is it important?
Real-world deployments of robot teams almost always involve heterogeneous fleets: different robots for different jobs, areas, or payloads. However, most existing algorithms assume all robots are identical and can go anywhere. This paper aims to formally study and solve the multi-robot pickup and delivery problem in these realistic settings. The framework enables practical coordination in scenarios like hospital logistics, automated supermarkets, and construction sites, and lays the groundwork for future work on learning-based coordination strategies and more complex robot heterogeneity (e.g., differing speeds).
Perspectives
This work is motivated by the observation that in many real-world applications, (robotic) agents are not homogeneous and are required to collaborate to solve tasks. Formalizing such heterogeneity within the MAPD framework and designing a corresponding planning approach addresses a gap that has received limited attention so far. We hope this work contributes to a broader shift toward approaches for coordinated multi-agent path planning that explicitly account for the inherent diversity of real-world robotic teams.
Francesco Amigoni
Politecnico di Milano
Read the Original
This page is a summary of: Multi-Agent Pickup and Delivery with Heterogeneous Agents, International Foundation for Autonomous Agents and Multiagent Systems,
DOI: 10.65109/gkgu6726.
You can read the full text:
Contributors
The following have contributed to this page
