Article: IMAP-QL: an improved multi-agent pursuit path-planning based on Q-learning Journal: International Journal of Systems, Control and Communications (IJSCC) 2024 Vol.15 No.2 pp.159 - 178 Abstract: Multi-agent path planning is a complex problem aiming to find the shortest trajectory. In this paper, we propose a cooperative path planning based on a hybridisation of two motion behaviours. The first behaviour concerns the strategy in which the agents are independently moving to their common objective according to the detected environment's rewards. The second behaviour is based on the leader-followers strategy through which the follower agents are moving in the direction of their leader agent to achieve their objectives. The goal of this work is to provide an equilibrium between these two approaches by allowing the follower agents to move in the direction of the objective instead of following the leader in some cases to decrease the execution time. To prove the approach's feasibility, we applied it to the multi-pursuer multi-evader game in comparison with the recent approaches. The obtained results prove the efficiency of the proposed approach. Inderscience Publishers - linking academia, business and industry through research

Title: IMAP-QL: an improved multi-agent pursuit path-planning based on Q-learning

Authors: Mohammed El Habib Souidi; Makhlouf Ledmi; Toufik Messaoud Maarouk; Abdeldjalil Ledmi; Ferial Laassami

Addresses: Department of Computer Science, ICOSI Lab, University of Khenchela, 40004 Khenchela, Algeria ' Department of Computer Science, ICOSI Lab, University of Khenchela, 40004 Khenchela, Algeria ' Department of Computer Science, ICOSI Lab, University of Khenchela, 40004 Khenchela, Algeria ' Department of Computer Science, ICOSI Lab, University of Khenchela, 40004 Khenchela, Algeria ' Department of Computer Science, ICOSI Lab, University of Khenchela, 40004 Khenchela, Algeria

Abstract: Multi-agent path planning is a complex problem aiming to find the shortest trajectory. In this paper, we propose a cooperative path planning based on a hybridisation of two motion behaviours. The first behaviour concerns the strategy in which the agents are independently moving to their common objective according to the detected environment's rewards. The second behaviour is based on the leader-followers strategy through which the follower agents are moving in the direction of their leader agent to achieve their objectives. The goal of this work is to provide an equilibrium between these two approaches by allowing the follower agents to move in the direction of the objective instead of following the leader in some cases to decrease the execution time. To prove the approach's feasibility, we applied it to the multi-pursuer multi-evader game in comparison with the recent approaches. The obtained results prove the efficiency of the proposed approach.

Keywords: multi-agent system; path planning; reinforcement learning; pursuit evasion game.

DOI: 10.1504/IJSCC.2024.138549

International Journal of Systems, Control and Communications, 2024 Vol.15 No.2, pp.159 - 178

Received: 02 Dec 2023
Accepted: 14 Feb 2024
Published online: 10 May 2024 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article

Title: IMAP-QL: an improved multi-agent pursuit path-planning based on Q-learning

Keep up-to-date