State Estimation subject to Packet Losses

[June 2016 - Dec 2017]

The project aimed at studying state estimation for Linear Time Invariant (LTI) systems, subject to random packet dropouts. Addressing different ways to incorporate this randomness led to two different sub-problems:

(i) packet losses that are Markovian in nature: these packet losses are common in communication channels (such as ethernet) and are used to model probabilistic packet droupouts occurring across these channels

(ii) packet losses that occur within specific regions, where the regions are themselves distributed according to known priors. These can model situational awareness by the LTI system of the presence of radio jammers and other adversarial entities.

By systematically utilizing the a priori information of the regions where the packet loss is likely to occur, the proposed estimator takes the Kalman filter structure with the modified algebraic Riccati iteration for the error covariance matrix being stochastic due to the probabilistic packet arrival process. This to an improved accuracy of the proposed filter compared with the baseline Kalman filter, with a comparable computation complexity.

Principal Investigators list

• Omanshu Thapliyal

• Jayaprakash S. Nandiganahalli

Related Publications

• O. Thapliyal, J.S. Nandiganahalli and I. Hwang, “Kalman Filtering with State-Dependent Packet Losses”, IET Control Theory and Applications, Vol.13(2), pp.306-312, February 2019, DOI: 10.1049/iet-cta.2018.5425

• O. Thapliyal, J.S. Nandiganahalli, and I. Hwang, "Optimal State Estimation in LTI Systems with Imperfect Observations," In the Proceedings of the 56th IEEE Conference on Decision and Control, Melbourne, Australia, December 2017