Intelligent Resource Management In Internet Of Things For Multiple Targets Tracking

As an important kind of Io T,Multiple Target-Tracking Wireless Sensor Network(MTT-WSN)has significant advantages of low energy consumption and easy deployment.However,it cannot achieve real-time target sensing due to limited sensing resources for real-time target tracking.MTT-Io T can introduce flexible UAV networks further to expand the ranges of target sensing and tracking.Unfortunately,UAVs cannot effectively predict target trajectories with limited computing resources for real-time MTT.Moreover,it is difficult to perform effective information exchange among UAVs with scarce communication resources.With considerations of high dynamic of targets,UAVs cannot also ensure effective path planning for accurate MTT.To address the mentioned challenges,this dissertation proposes a UAV swarm-based MTT-WSN tracking framework to ensure real-time and accurate MTT.The work of this dissertation can be divided into four parts:1)To address the issue of the limitation of sensing resources in small-scale MTT scenarios,this dissertation proposes a hybrid WSN-based intelligent sensing cooperation strategy.It can enable mobile WSN nodes to dynamically activate surrounding static WSN nodes for cooperative target sensing.Furthermore,a dynamic resource scheduling algorithm is proposed to integrate computing resources of mobile WSN nodes and edge servers for acquiring accurate trajectory prediction results and cooperative tracking decisions.In this case,our solution guarantees real-time target sensing and tracking.2)Considering high computing overhead for tracking high-speed moving targets in middle-scale scenarios,this dissertation proposes an Unmanned Aerial Vehicles(UAV)-based computing resource cooperative management scheme.It can assist UAVs in acquiring information of neighbors to select suitable neighbors for cooperative trajectory prediction with low-latency prediction performance.To improve the accuracy of trajectory prediction,the scheme combines the advantages of the EKF algorithm and PSO algorithm to accelerate the prediction process for real-time trajectory prediction for accurate tracking.In addition,a UAV swarm-based intelligent tracking cooperation algorithm is proposed to optimize tracking paths for accurate MTT.3)In medium-scale MTT scenarios,cooperative computing can lead to high communication overheads.This dissertation proposes a Digital Twin(DT)based cooperative scheduling of sensing and communication.It can extract information of UAVs and targets accurately from physical MTT scenarios using an attention scheme for building accurate DT models.The models can predict and derive mobile trajectories and velocities of UAVs and targets to select feasible UAVs for cooperative target tracking based on the optimization of the DDPG framework.For targets with similar speeds to that of UAVs,the models can select suitable neighbors to perform accurate cooperative tracking.For targets with high speeds,the models can conduct UAVs to invite remote UAVs for cooperative tracking.On the other hand,unlike traditional broadcast manner,the models can adjust beams of antennas to select feasible UAVs to reduce communication overhead by planning optimal transmission routes.4)DT cannot build real-time and accurate mapping with high computing complexity for large-scale MTT scenarios.This dissertation proposes a tiered DT-assisted tracking solution by optimizing the existing DT pattern.It adopts a dual-grained DT pattern to achieve real-time and accurate MTT.Unlike traditional terminal-edge-cloud architecture,the cloud server can predict and derive position relations between targets and UAVs.It can decompose the UAV swarm into multiple subgroups to associate and track feasible targets with a coarse-grained DT.Based on this,each subgroup leader implements a finegrained DT to derive sensing postures,mobile velocities,and trajectories of UAVs.With the derivation decision,UAVs can implement cooperative tracking accurately and in real time with the optimization of tracking paths.