Wireless Sensor Network Scheduling And Routing Optimization Based On Deep Reinforcement Learning

Wireless sensor networks are widely used in industrial production,intelligent transportation,environmental monitoring,etc.The main challenges are focused on realtime,energy management,deployment and positioning,routing,data fusion and compression,etc.,all aimed at solving Maximize the utility of wireless sensor networks.Reinforcement learning is one of the three branches of machine learning.The agent constantly interacts with the environment to learn to get the best strategy in a specific Scenes.This paper uses reinforcement learning to solve node scheduling and routing problems in wireless sensor networks.Node scheduling of wireless sensors is a hotspot of its research.How to select K nodes as active nodes in N nodes under the premise of satisfying coverage and detection probability,other nodes enter dormant state to save energy.,node scheduling method based on the Q-learning algorithm can update The scheduling strategy periodically,and update the Q value of the state-action pair through continuously learning.The actual scheduling strategy is output through the ?-greedy function during scheduling.Routing protocols have also been the hotspot research direction of wireless sensor networks.Network delay,load balancing,life time and energy balance are the problems that routing protocols need to solve.The routing path is planned based on the deep reinforcement learning algorithm-DQN.When the node local buffer is full,data is transmitted to the sink through multi-hop.The direct rewards of reinforcement learning mainly consider the step distance,single-hop energy consumption,energy balance of adjacent nodes,and the possible negative effects of retransmission.This paper uses Python to develop a simulation environment and an reinforcement learning training environment,and compares it with existing programs on related indicators.The node scheduling optimization experiment proves that this scheme achieves better results in coverage and energy balance than the random and shortest distance scheme.In the routing experiment,the experimental analysis of the Q value of the node proves the convergence of the algorithm,and compared with the protocols such as LEACH and FTIEE,it proves that the scheme can effectively extend the network life time and realize load balancing.