Research On Multi-source Sensor Resource Management And Task Scheduling Method

Multi-source sensor management supports the sensor network to complete tasks such as multi-target tracking and multi-target classification through resource allocation,and can improve the system response rate through task scheduling.This thesis studies sensor deployment and multi-task scheduling under multi-target tracking and multi-target classification,and proposes two algorithms and one model.The main work and results include:1.Aiming at the question of which strategy the sensor should adopt to select which targets to track in the multi-target tracking scenario,this thesis designs an estimation method based on fuzzy logic system to quantify the target threat,and further proposes a binary particle swarm optimization Algorithm(Binary Particle Swarm Optimizaion,BPSO)sensor allocation algorithm.A fuzzy logic system and uses Analytic Hierarchy Process(AHP)are designed to estimate the threat of the target from the four dimensions of the target's distance,speed,acceleration and heading angle.Considering that resources are tilted towards high-threat targets,a proposal A BPSO-based sensor allocation algorithm.In this thesis,a sensor selection algorithm based on BPSO is used to compare the utilization of sensors under different target numbers.The results show that the best sensor utilization is the highest in a multi-sensor system.2.Unlike the continuous-time multi-target tracking scenario,the multi-target classification problem has discrete characteristics,so it is difficult to conduct a unified performance evaluation and sensor deployment research on sensors.To solve this problem,this thesisproposes a measure to quantify the classification performance of sensors,and proposes a sensor deployment algorithm based on this measure.Based on the Fusion Performance Model(FPM)and other multi-object classification theories,this thesis proposes a symmetric single-sensor classification performance measurement,and conductes it to the case of multi-sensor combinations,and verifies its effectiveness through derivation and simulation.A sensor deployment algorithm is proposed based on the definition of the benefit ratio based on this measurement,and a multi-target classification network is constructed under the Un BBayes platform to verify the algorithm.The experimental results show that the deployment of sensors through the sensor deployment algorithm proposed in this thesis can balance energy consumption and improve at the same time The accuracy of multi-target classification.3.Unlike a single task which has a clear physical description,the task flow faced by a multi-sensor system in practice has characteristics such as diversity and randomness,so how to schedule tasks to meet the response speed of the system is an important research content.In order to meet the needs of multi-sensor systems to respond quickly to the system,this thesis proposes a task abstract model.Based on this model,two task ordering rules,High Type Early Deadline First(HTEDF)and Modified High Type Early Deadline First(MHTEDF)are given,and an adaptive multi-task scheduling algorithm is proposed on this basis for identity confirmation,high-precision tracking,and missing Target search,medium-precision tracking,low-precision tracking,and search are scheduled for six tasks.Under different number of tasks,the same sorting rule is used to compare the effect of genetic algorithm-based multi-task scheduling and adaptive scheduling algorithm on multi-task scheduling.The results show that the response speed of adaptive multi-task scheduling algorithm is half that of genetic algorithm.This thesis proposes a sensor allocation algorithm in the context of multi-target tracking,which can make full use of sensors to allocate more resources to high-threat targets.Furthermore,a sensor deployment algorithm is proposed for multi-target classification,and the obtained deployment plan can improve the accuracy of multi-target classification while balancing the energy consumption of sensors.Aiming at how to schedule the above tasks,this thesis proposes an abstract task model and gives an adaptive scheduling algorithm to improve the response speed of the system,which provides a theoretical basis for multi-source sensor management.