| Data collection is one of the important operations of wireless sensor network(WSN).The traditional data collection method is mainly based on multi-hop forwarding,which is easy to cause energy holes around the sink node.Subsequently,some studies introduced mobile robots or cars to collect data,so as to avoid energy holes.However,these robots or cars move slowly and are greatly affected by terrain and obstacles.Recently,unmanned aerial vehicle(UAV)has attracted extensive attention from scholars due to its high mobility,flexible deployment and low cost,and the UAV-based WSN data collection has also become one of the research hotspots.This thesis introduces the concept of value of information(VoI)and employs it as an index to evaluate the performance of data collection.For single UAV,multiple UAVs,and large-scale scenarios,the maximum VoI first and successive convex approximation(MVF-SCA),balanced allocation and K-means(BA-KM)and direct future prediction(DFP)model based trajectory planning algorithms are proposed respectively.At last,the performance of the proposed algorithms are analyzed by simulations.The main contributions of the thesis are as follows.(1)For the single UAV data collection scenario,the MVF-SCA algorithm is proposed with the goal of maximizing the VoI of data collected by the UAV.First,the optimization problem is modeled as a mixed integer nonlinear programming problem.Then,the MVF-SCA algorithm is proposed to find suboptimal solutions with low complexity.Finally,the proposed algorithm is simulated.The results show that the VoI collected by the MVF-SCA algorithm is about 15%to 30% higher than the classic Traveling Salesman Problem(TSP)algorithm and Greedy Path Planning(GPP)algorithm.(2)For the multipe UAVs data collection scenario,the BA-KM algorithm is proposed with the goal of maximizing the VoI of data collected by all UAVs.The algorithm divides all nodes into multiple sets without intersection,and each node set sends a UAV to collect data.Simulation results show that,for different node distributions,the performance of BA-KM algorithm is more stable than that of the equal division algorithm and the classical K-means algorithm.(3)For the large-scale WSN scenario,the algorithm based on DFP model is proposed with the goal of maximizing the total VoI collected and ensuring the timely charging of the UAV when the power is low.First,the large scale WSN is clustered in a grid to select cluster head nodes and determine the forwarding strategy of nodes within the cluster.Then,the algorithm based on DFP model is proposed,in which only the cluster-head nodes are collected by the UAV.Finally,the proposed algorithm is compared with Q-Learning algorithm and Deep QNetwork(DQN)algorithm.The simulation results show that the proposed algorithm has better collection performance than Q-Learning algorithm and DQN algorithm. |
