Research On Mobile Energy Supplement In Wireless Sensor Networks Based On Location Selection Mechanism And DR

With the development of sensing technology,wireless sensor network(WSN)with low deployment cost and easy scalability has been widely used.However,due to limited sensor battery capacity and variable energy consumption of nodes,failure of some nodes due to energy depletion may lead to overall communication failure of the network.Wireless charging technology can effectively solve the problem of energy limitation of WSN by using Mobile Charger(MC)to charge the sensor nodes,which ensures the stable operation of WSN for a long time.Most of the current research works on energy replenishment based on one-to-many wireless charging technologies have been conducted experimentally by pre-planning MC charging paths with the assumption of infinite MC energy.They do not consider the existence of dynamic changes in the network with limited MC energy,ignore the charging cost and do not incorporate the actual production environment to achieve energy replenishment.Therefore,this paper divides the energy replenishment problem into two sub-problems of stationing point selection and path planning to solve them separately in order to address the shortcomings in previous work,and the following two optimization methods are proposed:(1)A one-to-many charging strategy based on set coverage problem with DDQN,referred to as MSRL,is proposed for the high energy replenishment demand in sensor network application scenarios.Firstly,MSRL abstracts the MC charging stationing point selection problem as a set coverage problem with weights,and solves the approximate optimal set of stationing points based on WSC＿RA algorithm;secondly,the environment is modeled based on markov chain,and the comprehensive considering the energy consumption rate,geographical location and remaining energy of sensor nodes,adaptively adjusting the charging path using the Dueling DQN algorithm,and further accelerating the network training speed using the Gradient Bandit strategy to achieve the optimization of MC scheduling;finally,multiple simulation experiments were guided and the experimental results showed that the MSRL approach can significantly reduce the number of sensor node deaths,increase the average energy value of the network and extend the survival time of the network,and outperform its comparative approach.(2)Considering the optimization of the charging utility of the mobile charging device,so that more energy is used to charge the sensor nodes,a full-coverage path mobile charging strategy based on deep reinforcement learning,referred to as MNCC,is proposed with the goal of optimizing the distance traveled by the mobile charging device.Firstly,MNCC adopts an improved dichotomous K-mean algorithm to cluster the network and realize the replenishment of energy for nodes in clusters;then,considering the dynamic variability of charging requests,the MC charging path is constructed based on the pointer network model,and the neural network parameters are trained using the policy gradient algorithm to realize the mapping from problem sequence to solution sequence;finally,a large number of simulation experiments are guided,and the results show that the method can effectively reduce the MC moving distance and node failure rate,improve the MC charging efficiency,and thus achieve the network survival time and outperforms its comparative methods.