Introduction
Bimonthly, started in 1957
Administrator
Shanxi Provincial Education Department
Sponsor
Taiyuan University of Technology
Publisher
Ed. Office of Journal of TYUT
Editor-in-Chief
SUN Hongbin
ISSN: 1007-9432
CN: 14-1220/N
Administrator
Shanxi Provincial Education Department
Sponsor
Taiyuan University of Technology
Publisher
Ed. Office of Journal of TYUT
Editor-in-Chief
SUN Hongbin
ISSN: 1007-9432
CN: 14-1220/N
location:
home> Online First
DOI:
10.16355／j.tyut.1007-9432.20230016
abstract:
Reinforcement learning policy transfer is an effective way to reduce the consumption of deep reinforcement learning training.Local policy transfer is policy transfer at a fine-grained, which is of great significance to the improvement of the global policy performance and the formation of a new global policy by the combination of local policy. However, most current researches on policy transfer focus on the global policy transfer, and relatively few research on local policy transfer. Therefore, this paper proposes a deep reinforcement learning method for local policy transfer. This method draws on the idea of "high cohesion, low coupling" in software engineering. By dividing the neural network, which is the carrier of policy, different sub-neural networks carry different local policies, and then realizes the transfer of local policies through the transfer of sub-neural networks. This method supports flexible replacement and combination of local policy and forms a new global policy with better performance and adaption to new environment. In this paper, the classical deep reinforcement learning algorithm DQN is selected as the experimental algorithm and the transfer ability and performance of DQN algorithm before and after using the proposed method is compared. The results show that the DQN algorithm realizes local policy transfer and improve the performance by about 27.5% after using the proposed method.
Keywords:
Deep reinforcement learning; local policy transfer; DQN