Abstract: To overcome the serve energy consumption problem in indoor wireless networks, a deep Q-Learning based transmit power allocation algorithm for home base station is proposed. Firstly, a deep learning network (DLN) is built to optimize the energy efficiency of indoor wireless networks. Then, the energy consumption rating is regarded as the rewards, the batch gradient descent method is used to continuously train the weights of DLN. Finally, the simulation results show that the proposed algorithm can dynamically adjust the transmit power, and is significantly superior to Q-Learning and water-filling algorithms in terms of convergence speed and EC optimization, which can effectively reduce the energy consumption of indoor wireless networks.

Key words: indoor wireless network, energy efficiency, power allocation, deep Q-Learning