Skip to Main content Skip to Navigation
New interface
Journal articles

Decentralized Multi-Agent Pursuit Using Deep Reinforcement Learning

Abstract : Pursuit-evasion is the problem of capturing mobile targets with one or more pursuers. We use deep reinforcement learning for pursuing an omnidirectional target with multiple, homogeneous agents that are subject to unicycle kinematic constraints. We use shared experience to train a policy for a given number of pursuers, executed independently by each agent at run-time. The training uses curriculum learning, a sweeping-angle ordering to locally represent neighboring agents, and a reward structure that encourages a good formation and combines individual and group rewards. Simulated experiments with a reactive evader and up to eight pursuers show that our learning-based approach outperforms recent reinforcement learning techniques as well as nonholonomic adaptations of classical algorithms. The learned policy is successfully transferred to the real-world in a proof-of-concept demonstration with three motion-constrained pursuer drones.
Document type :
Journal articles
Complete list of metadata
Contributor : Pedro Castillo Garcia Connect in order to contact the contributor
Submitted on : Monday, November 7, 2022 - 5:04:57 PM
Last modification on : Wednesday, November 23, 2022 - 3:19:13 AM

Links full text



Cristino de Souza, Akansel Cosgun, Pedro Castillo Garcia, Boris Vidolov, Dana Kuli. Decentralized Multi-Agent Pursuit Using Deep Reinforcement Learning. IEEE Robotics and Automation Letters, 2021, 6 (3), pp.4552-4559. ⟨10.1109/LRA.2021.3068952⟩. ⟨hal-03842721⟩



Record views