Event-Based Information Acquisition, Learning, and Control in High-Dimensional Cyber-Physical Systems

Abstract:

This research proposal focuses on event--based information acquisition, state estimation and control in the context of high dimensional cyber physical systems. In particular, as part of the cyber system a (set of) decision maker(s) or agents is responsible for the acquisition of information, learning, and control about the underlying physical system of interest. The information acquisition process may be instigated or adapted based on events in the systems. We consider sample collection and information acquisition in the form of both a pull and push strategy. The main challenge in the pull strategy is to enable fast response and accurate centralized tracking of stochastically varying parameters in the face of constrained measurements (in terms of dimensions or noise characteristics). As an alternative to centralized measurements, distributed information acquisition has to address the issues of scalability and in particular the impact of decentralization of decision making on the performance. In such scenarios, also referred to as push strategies, the task of event--driven sample collection is leD as a local decision to the sensors. Motivated by a synthesis of the PIs' (independent and collaborative) prior works on (distributed) sensing, communications, and control as well as cross--layer network design for control applications, this framework is particularly apt for problems of sensing and associated control in high dimensional systems. We plan to close the loop by considering estimation and control jointly over these event--based information acquisition policies. We will verify whether a separation principle between information acquisition/learning and control might hold, and also develop the optimal estimator and controller for such scenarios. Intellectual Merit: The intellectual merit of this work will be to develop a theoretical framework for the design of cyber--physical systems including information acquisition, learning, and control. Separation theorems for the optimality of separate state estimation and control will be explored. These results will provide insight as well as significantly improved performance for the next generation of control systems with distributed sensing. Broader Impacts: Significant performance improvement of event--based information acquisition, state estimation and control in the context of high dimensional cyber physical systems will improve the performance for a wide range of applications for societal benefit, including smart buildings, intelligent transportation, energy--efficient data centers, and the future smart--grid. The PIs plan to disseminate the research results from this project through presentations and papers at high quality conference and journal venues, as well as by organizing specialized workshops and conference sessions related to the application of stochastic learning to the control of cyber physical systems. The proposed project will aid in the training of Ph.D. students as well as enrich the curriculum of courses taught by the PIs in communications, stochastic control, and networks. The PIs, which include 2 women, have a strong track record in outreach and contributions to diversity and will also incorporate the results and activities of this research in their outreach activities with high school and undergraduate students, including under-- represented minorities and women.