Jump estimators for lossy wireless structural control
The implementation of wireless sensor networks for structural control may offer advantages over a wired network,including reduced installation and maintenance cost and the possibility of a wide variety of network architectures.However,wireless transmission involves signal loss,which may contribute to a decrease in system performance.For linear systems,a time-varying Kalman filter(TVKE)provides optimal estimation in the presence of measurement losses.However,a TVKE requires solving a Riccati equation and re-computing the estimator gain after each time step.On the other hand,a jump linear estimator(JLE)switches between a finite set of pre-computed modes,which replaces the task of re-computing the gain at the sacrifice of some performance.This work proposes the use of a specific class of JLE,the finite-loss-history estimator(FLHE),which selects the estimator gain based on previous finite-length lost-received sequences.Previous work established the optimal solution for FLHE for systems where measurement losses are model as a Markovian process.In this study,FLHEs are extended to civil engineering applications and applied to an estimation problem on a three-story benchmark structure.The estimator receives measurements of floor accelerations,memorizes and uses the loss history for gain switching,and estimates the fullstate feedback for an optimal quadratic regulator.The number of pre-computed estimator modes increases with the finite length of the memorized loss history sequence.Taking more time steps into the history sequence is expected to give better performance; however,the level of improvement will be dependent on the input time-history,which can be highly transient and uncertain for civil systems.Simulations are conducted to compare the performances of TVKE and FLHEs with different history lengths.Ultimately,the goal is to determine a loss history length that balances performance,the number of estimator modes,and sensitivity to the input.
jump linear estimator signal loss finite-loss-history estimator wireless control Markovian jump system
T.H.T.Truong P.Seiler L.E.Linderman
PhD Student,Department of Civil Environmental,and Geo-Engineering,University of Minnesota,Twin Citie Associate Professor,Department of Aerospace Engineering and Mechanics,University of Minnesota,Twin C Assistant Professor,Department of Civil Environmental,and Geo-Engineering,University of Minnesota,Tw
国际会议
The 7th World Conference on Structural Control and Monitoring(7WCSCM)(第七届结构控制与监测世界大会)
青岛
英文
808-817
2018-07-22（万方平台首次上网日期，不代表论文的发表时间）