The goal of this project is to develop concepts and tools for model-based development of embedded control systems for automotive x-by-wire systems. X-by-wire systems replace mechanical connections between the driver and the vehicle with embedded computers and actuators (e.g., motors) to control the automobile and provide physical responses to the driver. These systems are characterized by multiple, highly integrated and interacting feedback loops. This project is creating new methods for designing these embedded control systems to improve performance while enhancing safety and reliability and reducing development cost. The research draws on control theory to deal with physical dynamics and synthesis of the feedback loops and with formalisms and techniques from computer science to deal with the logical aspects of the embedded software. Using a laboratory implementation of a steer-by-wire system as a test bed, models of the dynamics, including driver behavior, are being analyzed using new compositional methods that make it possible to design elements of the embedded control systems with guaranteed performance characteristics. The theory of fundamental design limitations is being extended and applied to these models to evaluate the tradeoffs between dynamic performance and robustness. Verification methods for hybrid systems (systems with continuous and discrete state variables) are being developed to demonstrate the safety of the system, even when there are failures in sensors and other system components. Tractable verification problems are being developed based on model decompositions reflecting the model structure, time-scale separation, and novel representations of the interaction dynamics. The embedded software design is being implemented and evaluated on the laboratory test bed in cooperation with an industrial partner.
