Reinforcement Learning Based Content Push Policy for HetNets with Energy Harvesting Small Cells.

In order to utilize renewable energy and save traditional energy, many literatures in recent years have drawn attention to content caching in wireless communications. In this article, we focus on content push and cache to increase green energy utilization and save traditional energy. The state transition probability and future rewards in the mobile environment are unknown. Therefore, we use reinforcement learning to solve the problem of green energy distribution and the content push. Q-Learning is a model-free enhanced learning technology that can find an optimal action selection strategy in the MDP question. The Boltzmann distribution method is used to update the strategy. Finally, we can find the desired action based on the current state and the optimal strategy. SBS selects actions according to the Boltzmann strategy and then iteratively updates the Q-tables to get the best action in each state. Through numerical simulation, we prove the validity of the model and get the regularity of SBS’s decision.