Optimal sampling and switching policies for linear stochastic systems

For systems with limited capacity for storage, processing and transmission of data, the choice of sampling policy is critical. Although most systems determine their sampling instants in advance, for example periodically, this results in unnecessary use of samples if little changes occur between sampling times. Instead, to optimize the utilization of the samples, the decision to take a sample can be adaptively made based on the importance of the change in the state of the system. This calls for development of event-triggered sampling policies. In this thesis, we study the optimal event-triggered sampling policies under a constraint on the frequency of sampling. We first investigate the optimal sampling policies to minimize the estimation error over the infinite horizon. The optimal policies are provided for multidimensional Wiener processes and scalar linear diffusion processes. Then, we address an infinite horizon control problem with a stochastic process driven by a bang-bang controller. We obtain the optimal times to switch the control signal that determines the drift rate of the process. For the cases handled in this thesis, the results suggest the optimality of the simplest event-triggered sampling policy with constant thresholds over the infinite horizon.

Name thesis:degree_name

M.S.

Level thesis:degree_level

Thesis

Discipline thesis:degree_discipline

Electrical & Computer Engr

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2014

Author dc:creator

• Nar, Kamil

Contributors dc:contributor

• Basar, Tamer

• Event-triggered sampling
• Optimal stopping

Statement dc:rights

• Copyright 2014 Kamil Nar

Language dc:language

en

Handle dc:identifier

http://hdl.handle.net/2142/50571

OAI identifier oai:identifier

oai:www.ideals.illinois.edu:2142/50571