基于深度强化学习的应急通信网络规划方法

针对应急通信网络规划传统算法对先验知识要求高、时效性不强等问题,提出一种基于深度强化学习的应急通信网络拓扑规划方法。研究了基于蒙特卡罗树搜索与自博弈相结合的网络规划样本数据生成方法,设计了基于残差网络的策略网和价值网,在此基础上使用Tensorflow库对模型进行构建和训练。仿真结果表明,提出的规划方法能够有效实现网络拓扑的智能规划,且具有较高的时效性和可行性。

Emergency communication network planning method based on deep reinforcement learning

Focus on the problem of high demand on prior knowledge and weak timeliness of traditional algorithm for emergency communication network planning, a toplogy planning method for emergency communication network based on deep reinforcement learning is proposed. Developing a method of sample data generation using Monte Carlo tree search and self-game, the policy network and value network based on residual network is designed. On this basis, Tensorflow is used to build and train the model. Simulation results show that the proposed planning method can effctively realize the intelligent planning of network topology, and has high timeliness and feasibility.