Distributed Estimator Design For Formation Systems

With the rapid development of communication technology and controltheory, distributed estimator technology came into being and in the field ofenergy system and sensor networks get more widely used. At the same time，in the field of aerospace, the multi-agent formation system can achieve morecomplex task, the theory and application of distributed estimator not onlyprovided for the formation system of the global information, but also toimprove the formation speed.For a formation of agent system can be described as discrete lineartime-invariant system，research on design method for distributed estimationin this article. First introduces the research background and current status ofdistributed estimator, after Propose the distributed estimator model based onexisting research. Finally, according to the characteristics of the differentmodels using linear matrix inequalities, Markov process, the stateaugmentation design different algorithms.Assume that each agent has local estimates, you can estimate of theentire formation state. Each estimator has the communication with others.Communication process use different models with different environment anddesign different algorithms. According to the actual situation we research four communicationmodels:1. The ideal communication model: Communication process is nopacket dropout and noise problems. By solving the linear matrix inequalitiesof the system error covariance matrix, access to estimator filter gain andtransmitter gain.2. Communication model with packet dropout: The communicationpacket dropout process is a Markov process, using different Markov stateconstituted by the system error covariance matrix. It can obtain a feasiblesolution. Ultimately, we can get parameters of the estimator.3. Communication model with noise and bandwidth restrictions: Thecommunication process has noise interferencean, and communicationbandwidth is limited. By using matrix equivalent transformation of bilinearmatrix inequalities into two linear matrix inequalities, get estimator filtergain and transmitter gain.4. Communication model with noise and packet dropout: Summary ofthe preceding two methods to achieve the design of distributed estimation.Finally, Study various factors affecting the performance of distributedestimator.