A quantum inspired reinforcement learning technique for beyond next generation wireless networks

摘要

This paper proposes the application of a quantum inspired reinforcement learning technique for spectrum assignment of wireless communication networks. The proposed technique aims to enhance the speed of learning convergence through the dependence of the decision process on a well ranked action desirability table which is updated based on the success or failure of an action. In addition, the exploration process is exclusively induced by the failure of the channel choice and directs the agent to the next best channel. The quantum technique is compared with traditional reinforcement learning, random assignment reinforcement learning, and random dynamic channel assignment algorithms. This quantum technique is shown to increase the speed of learning convergence of traditional reinforcement learning by up to 40 times. Thus, system capacity can be improved in terms of the number of users by (9-84) %, and provides a significant average file delay reduction of 26% on average, and throughput improvement of up to 2.8%.