Data Driven Intelligent Controlled Sensing for Cyber Physical Systems
Abstract:
Many cyber-physical systems (CPS) deployed in a number of applications ranging from airport security systems and transportation systems to health-care and manufacturing rely on a wide variety of sensors for prediction and control. In many of these systems, acquisition of information requires the deployment and activation of physical sensors, which can result in increased expense or delay. A recent explosion in sensor and UAV technology has lead to new capabilities for controlling the nature and mobility of sensing actions by changing excitation levels, position, orientation, sensitivity, and similar parameters. This has in turn created substantial challenges to develop cyber-physical systems that can effectively exploit the degrees of freedom in selecting where and how to sense the environment. These challenges include high dimensionality of sensor observations, poor understanding of models relating the nature of potential sensing actions and the corresponding value of the collected information, and lack of sufficient training data from which to learn these models. This project will contribute to the Science of CPS and provide concepts and design tools for intelligent operation of cyber-physical systems. We will synergistically integrate tools from multiple disciplines including distributed sensing, machine learning, stochastic control, and mobile sensor navigation to develop concepts for adaptively controlling information acquisition through the choice of sensor configurations to interact with the physical environment. Our goal is on a novel aspect of cyber-physical systems, intelligent sensing, to control information acquisition under sensing budget constraints in support of critical decision tasks. A novel data-driven approach to intelligent operation of cyber-physical systems is proposed to account for scenarios where models for sensing, decision-making, and prediction are unavailable or poorly understood. We propose multi-stage decision systems to incorporate sensor mobility, delays, costs, uncertainties, and risks into decision-making.
Research is proposed along three major thrusts:
1) Systematic framework for understanding the inherent tradeoffs between risks, costs, and delays when underlying models are poorly understood.
2) Development of data-driven stochastic control methods for robustly optimizing costs when the underlying models are unknown.
3) Validation of control methods on a UAV sensor network in the real world domain of archaeological surveying.
The proposed research will impact the frontiers of CPS through integration of multiple disciplines including sensor management, statistical learning, and distributed sensing. The results of the project will be disseminated through conferences presentations, departmental seminars, journal papers, and special session organization. A number of workshops have been planned at premier conferences including ICML, Global- SIP, RSS and CDC. This project will also have an impact on the education of future engineers. Our plan is to integrate the advances generated under this grant into the classes of our curriculum, including course projects. Furthermore, research in this project will feature prominently in a new graduate level course at BU. Our research group has historically comprised researchers and students from a diverse range of ethnicities from both genders. Prof. Castanon is Boston University’s adviser for the Society of Hispanic Professional Engineers (SHPE), and will involve participation from SHPE students in the research, and in using the results of the research in outreach programs for disadvantaged high school groups as part of the outreach activities. Furthermore the grounding of this project in the application of archaeological surveying will ensure that concepts developed will have an impact more broadly on applications of societal relevance, including environmental monitoring, traffic surveillance, search & rescue, and security applications.