Distributed Average Tracking For Noisy Signals

Distributed average tracking is an emerging research direction in the research of multi-agent systems in recent years.For a multi-agent system,assuming that each agent has its own reference signal,the objective of distributed average tracking is to design a local control input for each agent by using only the information of local neighboring agents,so that all agents can track the average value of all time-varying reference signals.This thesis uses graph theory,matrix theory,probability theory,cybernetics and other related mathematical tools to study distributed average tracking of linear discrete-time multi-agent systems with control input delay and random packet dropout in the case where the reference signal is noisy.The main works and contributions are summarized as follows.First,we study the distributed average tracking problem of multi-agent systems with noisy reference signals,in which the primary objective is to design a distributed algorithm that can deal with noise,and attenuate the effect caused by the noise.For this purpose,we design a Kalman filter for each agent to estimate their reference signals,based on which we design the DAT algorithm.We prove that the proposed algorithm can finally achieve bounded tracking error even though the reference signals are noisy.Second,we study the distributed average tracking problem of multi-agent systems with control input delay and random packet dropout.Unlike most distributed average tracking studies,we use a state predictor for each agent to compensate the negative effects caused by the control input delay,analyze the communication network with packet dropout by introducing an expected communication network topology,and propose a simple multi-stage distributed average tracking algorithm,which is robust to initialization errors and can be traded off between communication/computation costs and steady-state tracking errors.We finally show that the control target of distributed average tracking can be obtained asymptotically in the sense of mean square under the proposed algorithm.