TY - GEN

T1 - Simultaneous input and state estimation of linear discrete-time stochastic systems with input aggregate information

AU - Yong, Sze Zheng

AU - Zhu, Minghui

AU - Frazzoli, Emilio

N1 - Funding Information:
V. CONCLUSION We presented filtering algorithms for simultaneous input and state estimation of linear discrete-time stochastic systems for systems in which the inputs are not completely known but some aggregate information of the unknown inputs is available, either as linear equality or inequality constraints. Using two complementary views of the partial input information, we study the properties of the filters including the conditions for stability and optimality in the minimum-variance unbiased sense. We show that the estimate error covariance of the filters decreases when input aggregate information is available in either form. Moreover, given an equality constraint, the estimates remain unbiased but when given as an inequality constraint, the estimates may be biased although the bias is imperceptible in our simulation example. ACKNOWLEDGMENTS This work was supported by the National Science Foundation, grant #1239182. M. Zhu is partially supported by ARO W911NF-13-1-0421 (MURI) and NSF CNS-1505664. REFERENCES
Funding Information:
The work of X. Xu was supported by NSF grant CNS-1239037. The work of N. Ozay was supported in part by NSF grant CNS-1446298. The work of V. Gupta was supported in part by NSF grant CNS-1035655.

PY - 2015/2/8

Y1 - 2015/2/8

N2 - In this paper, we present filtering algorithms for simultaneous input and state estimation of linear discrete-time stochastic systems when the unknown inputs are partially known, i.e., when some aggregate information of the unknown inputs is available as linear equality or inequality constraints. The stability and optimality properties of the filters are presented and proven using two complementary perspectives. Specifically, we confirm the intuition that the partial input information improves the performance of the filters when a linear input equality constraint is given. On the other hand, given a linear inequality constraint, we show that the estimate error covariance is decreased but the estimates may be biased.

AB - In this paper, we present filtering algorithms for simultaneous input and state estimation of linear discrete-time stochastic systems when the unknown inputs are partially known, i.e., when some aggregate information of the unknown inputs is available as linear equality or inequality constraints. The stability and optimality properties of the filters are presented and proven using two complementary perspectives. Specifically, we confirm the intuition that the partial input information improves the performance of the filters when a linear input equality constraint is given. On the other hand, given a linear inequality constraint, we show that the estimate error covariance is decreased but the estimates may be biased.

UR - http://www.scopus.com/inward/record.url?scp=84962022983&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84962022983&partnerID=8YFLogxK

U2 - 10.1109/CDC.2015.7402243

DO - 10.1109/CDC.2015.7402243

M3 - Conference contribution

AN - SCOPUS:84962022983

T3 - Proceedings of the IEEE Conference on Decision and Control

SP - 461

EP - 467

BT - 54rd IEEE Conference on Decision and Control,CDC 2015

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 54th IEEE Conference on Decision and Control, CDC 2015

Y2 - 15 December 2015 through 18 December 2015

ER -