Deep Reinforcement Learning for antennas positioning

In this paper we study the position planning of cellular antennas. Informally, we try to find the best positions of antennas by having the maximum coverage and the minimum cost. Several methods were used to solve this problem. Since we have two main objectives, metaheuristics, such as genetic algorithms, simulated annealing, particle swarm optimization, … were used. Actually, artificial intelligence, and more particularly deep learning are compared to other methods in this context. Our work is about applying reinforcement learning on this problem. In this vein, we tried to tune some reinforcement learning methods to solve this problem. We found that the SARSA agent with the EpsGreedy QPolicy is the best option for these kinds of problems.