Deep reinforcement learning-based topology optimization for self-organized wireless sensor networks

Wireless sensor networks (WSNs) are the foundation of the Internet of Things (IoT), and in the era of the fifth generation of wireless communication networks, they are envisioned to be truly biquitous, reliable, scalable, and energy efficient. To this end, topology control is an important mechanism to realize self-organized WSNs that are capable of adapting to the dynamics of the environment. Topology optimization is combinatorial in nature, and generally is NP-hard to solve. Most existing algorithms leverage heuristic rules to reduce the number of search candidates so as to obtain a suboptimal solution in a certain sense. In this paper, we propose a deep reinforcement learning-based topology optimization algorithm, a unified search framework, for self-organized energy-efficient WSNs. Specifically, the proposed algorithm uses a deep neural network to guide a Monte Carlo tree search to roll out simulations, and the results from the tree search reinforce the learning of the neural network. In addition, the proposed algorithm is an anytime algorithm that keeps improving the solution with an increasing amount of computing resources. Various simulations show that the proposed algorithm achieves better performance as compared to heuristic solutions, and is capable of adapting to environment and network changes without restarting the algorithm from scratch.