Multi-agent Service Area Adaptation for Ride-Sharing Using Deep Reinforcement Learning

Yoshida, Naoki; Noda, Itsuki; Sugawara, Toshiharu

doi:10.1007/978-3-030-49778-1_29

Cited by 5 publications

(1 citation statement)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The use of multi-agent systems (MAS) for complex and enormous tasks in real-world applications has recently attracted considerable attention. Examples include transport robots in automated warehouses [34], autonomous aircraft-towing vehicles [18], ridesharing services [12,36], office robots [31], and delivery systems with multiple drones [10]. However, simply increasing the number of agents may lead to inefficiency owing to redundant movements and resource conflicts such as collisions.…”

Section: Introductionmentioning

confidence: 99%

Standby-Based Deadlock Avoidance Method for Multi-Agent Pickup and Delivery Tasks

Yamauchi¹,

Miyashita²,

Sugawara³

2022

Preprint

Self Cite

View full text Add to dashboard Cite

The multi-agent pickup and delivery (MAPD) problem, in which multiple agents iteratively carry materials without collisions, has received significant attention. However, many conventional MAPD algorithms assume a specifically designed grid-like environment, such as an automated warehouse. Therefore, they have many pickup and delivery locations where agents can stay for a lengthy period, as well as plentiful detours to avoid collisions owing to the freedom of movement in a grid. By contrast, because a maze-like environment such as a search-and-rescue or construction site has fewer pickup/delivery locations and their numbers may be unbalanced, many agents concentrate on such locations resulting in inefficient operations, often becoming stuck or deadlocked. Thus, to improve the transportation efficiency even in a maze-like restricted environment, we propose a deadlock avoidance method, called standbybased deadlock avoidance (SBDA). SBDA uses standby nodes determined in real-time using the articulation-point-finding algorithm, and the agent is guaranteed to stay there for a finite amount of time. We demonstrated that our proposed method outperforms a conventional approach. We also analyzed how the parameters used for selecting standby nodes affect the performance.

show abstract

Section: Introductionmentioning

confidence: 99%

Standby-Based Deadlock Avoidance Method for Multi-Agent Pickup and Delivery Tasks

Yamauchi¹,

Miyashita²,

Sugawara³

2022

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

Cost-Aware Dynamic Task Sharing Among Decentralized Autonomous Agents: Towards Dynamic Patient Sharing Among Hospitals

Hayashi

2022

Agents and Multi-Agent Systems: Technologies and Applications 2022

View full text Add to dashboard Cite

Efficient Path and Action Planning Method for Multi-Agent Pickup and Delivery Tasks under Environmental Constraints

2022

View full text Add to dashboard Cite

We propose a method called path and action planning with orientation (PAPO) that efficiently generates collision-free paths to satisfy environmental constraints, such as restricted path width and node size, for the multi-agent pickup and delivery in non-uniform environment (N-MAPD) problem. The MAPD problem, wherein multiple agents repeatedly pick up and carry materials without collisions, has attracted considerable attention; however, conventional MAPD algorithms assume a specially designed environment and thus use simple, uniform models with few environmental constraints. Such conventional algorithms cannot be applied to realistic applications where agents need to move in more complex and restricted environments. For example, the actions and orientations of agents are strictly restricted by the sizes of agents and carrying materials and the width of the passages at a construction site and a disaster area. In our N-MAPD formulation, which is an extension of the MAPD problem to apply to non-uniform environments with constraints, PAPO considers not only the path to the destination but also the agents’ direction, orientation, and timing of rotation. It is costly to consider all these factors, especially when the number of nodes is large. Our method can efficiently generate acceptable plans by exploring the search space via path planning, action planning, and conflict resolution in a phased manner. We experimentally evaluated the performance of PAPO by comparing it with our previous method, which is the preliminary version of PAPO, the baseline method in a centralized approach, and fundamental meta-heuristic algorithms. Finally, we demonstrate that PAPO can efficiently generate sub-optimal paths for N-MAPD instances.

show abstract

Multi-agent Service Area Adaptation for Ride-Sharing Using Deep Reinforcement Learning

Cited by 5 publications

References 10 publications

Standby-Based Deadlock Avoidance Method for Multi-Agent Pickup and Delivery Tasks

Standby-Based Deadlock Avoidance Method for Multi-Agent Pickup and Delivery Tasks

Cost-Aware Dynamic Task Sharing Among Decentralized Autonomous Agents: Towards Dynamic Patient Sharing Among Hospitals

Efficient Path and Action Planning Method for Multi-Agent Pickup and Delivery Tasks under Environmental Constraints

Contact Info

Product

Resources

About