Title: Decision optimisation of a mobile robot in a complex environment based on memory sequence replay mechanism
Authors: Dongshu Wang; Kai Yang; Xulin Gao; Heshan Wang
Addresses: School of Electrical and Information Engineering, Zhengzhou University, Zhengzhou, Henan Province, China; Longmen Laboratory, Luoyang, Henan Province, China ' Henan Branch of China Mobile Communications Group Co. Ltd, Zhengzhou, Henan Province, China ' School of Electrical and Information Engineering, Zhengzhou University, Zhengzhou, Henan Province, China ' School of Electrical and Information Engineering, Zhengzhou University, Zhengzhou, Henan Province, China
Abstract: To build a behaviour decision method that can converge to stability in a complex environment, this paper constructs an intelligent learning algorithm which can quickly find the shortest path by simulating the memory replay in the hippocampus. During the environment identification, a dynamic top-k competition is designed so the agent can consider the influence of multiple obstacles. Meanwhile, a representation layer of the advanced characteristics is added to the algorithm, so the agent can adapt to the complex structural environment. In addition, simulating the memory playback in the hippocampus, through the reverse playback of memory sequence, the reward discount is introduced into the weight updating process of the motor neurons, making the agent understand the relation among the memory fragments and guarantee the global optimal decision. Physical experiments of a mobile robot show that the agent can find the shortest path, demonstrating the agent can make optimal decisions in a complex structural environment.
Keywords: behaviour decision; hippocampus; memory replay; off-task process; developmental network.
DOI: 10.1504/IJAAC.2023.133977
International Journal of Automation and Control, 2023 Vol.17 No.5, pp.461 - 486
Received: 24 Aug 2021
Accepted: 18 May 2022
Published online: 09 Oct 2023 *