The Design And Analysis Of Distributed Multi-agent Optimization Algorithms Under The Information Constraints

In recent years,the development of distributed optimization technology has attracted more and more scholars' attention.Its application gradually penetrates into various fields,including solving the problem of resource allocation in wireless networks,path planning,traffic control problems and so on.This paper mainly considered in wireless sensor networks,using network sensors to carry out real-time monitoring and acquire various data,then interacting with information from adjacent sensors to realize data fusion.The wireless sensor network here is a multi-agent system,and each sensor is an agent.In the case of a non-smooth cost function for each sensor,the objective function is the sum of the cost functions for all sensors.The usual optimization method is using sub-gradients,but for non-smooth objective functions,it is very difficult or even impossible to calculate sub-gradients.In this thesis,the gradient-free algorithm is used to analyze and solve the problem that the objective function of the network is nonsmooth,but the objective function is Lipschitz continuous.The constrained optimization problem is solved to ensure that the estimate of the agent converges to the average value and the objective function converges to the optimal value.This brief description of the work is as follows:1.Considering that the connection between all the individuals in the wireless sensor network is fixed,the distributed random gradient-free algorithm is used to solve the wireless sensor network optimization problem.Due to the limitation of communication bandwidth between adjacent agents,quantization is used to effectively transmit information.Only the accuracy of quantization is considered here.Analyze the convergence of the state values of all agents in the network and the approximate convergence of the objective function to the optimal value through iteration.2.Based on the first point,the connection between agents in wireless sensor network is time-varying.Considering the influence of quantization accuracy,the state of agent in the network and the convergence of the objective function are also analyzed.3.Consider the actual situation,when communicating between neighboring agents,there are specific quantitative requirements.Therefore,the design of a uniform quantizer,quantization error is related with quantization interval and the communication bandwidth,so in the estimate of agent and the convergence of objective function analysis,the impacts of the quantization interval and the communication bandwidth need be considered.