Integrated System-Theoretic and Software-Enabled Strategies for Embedded Smart Sensors in Hybrid Systems

Georgia Tech Research Corporation, Atlanta GA

Investigators

Abstract

Bonnie Heck CCR-0209179 "Integrated System-Theorectic and Software-Enabled Strategies for Embedded Smart Sensors in Hybrid Systems" This research addresses the efficient design and effective operation of "smart" sensor suites, which are to be embedded in complex hybrid systems with multiple modes of operation and multiple mission / fault scenarios. An integrated approach is being pursued in which system-theoretic analysis is combined with real-time embedded software design techniques that focus on composition and dynamic reconfiguration issues. This integrative approach supports the design of the sensor suite, including its management software, which is capable of fusing information from heterogeneous sensors to develop estimates for physical parameters and detect / diagnose faults, and of dynamically reconfiguring itself in the face of identified fault scenarios or mode changes. The research addresses system-theoretic issues and software-design issues on a concurrent basis so that the requirements and capabilities of the two sides can be understood and taken into account. Hence, issues such as synchronizing data from multiple distributed sources and managing transitions while reconfiguring continuous components, are being formalized and explicitly accounted for throughout the design, beginning with the sensor selection / placement decision to the development of algorithms for sensor fusion and dynamic reconfiguration. This work is drawing upon existing results in the hybrid systems literature and developing some additional theories and tools as necessary. The topics studied in this research include: o Sensor placement in a distributed hybrid system: What types of sensors should be placed where so that important changes, both continuous and discrete, can be observed with required accuracies. o Sensor fusion: How the signals generated from a heterogeneous sensor array will be combined to generate useful information, such as physical states, parameter values, and fault status, under different modes of operation. o Dynamic reconfiguration: How to reconfigure on the fly the processing and fusing of signals from different sensors, in the face of an identified fault or a mode change, and how the information will be transitioned from one mode to another. The research is creating system-theoretic and software modeling tools for constructing smart sensor suites that can adapt to changes and faults that occur in a complex hybrid system. The research is also expected to improve the fundamental understanding into the observability property of hybrid systems and the design and analysis of a state estimator for such a system. The theoretical / methodological development is being validated on sensor suites for real applications (e.g., a three tank process demonstrator) so that no important real issues are overlooked. On the educational front, cross-disciplinary course(s) and research programs are being developed that breed a unique set of students versed in both systems-control theories and real-time software design.