| Abstract
|
Software is used to control complex system … Software is used to control complex systems as varied as spacecraft,land vehicles, life support systems, and factories. Traditionalmethods for developing control software do not support theflexibility, autonomy, and reliability required of these systems,and often do not support the hierarchical composition of such systems from independently developed components. For many of theseinherently hybrid (discrete and continuous) systems, combiningtraditional feedback PID control with discrete computer-basedcontrollers is a significant challenge in the design andimplementation of the control systems. This project proposes a newparadigm, model-enabled control, that uses declarative modelsaugmented with automatic reasoning systems to support the design,development, implementation, and validation of controllers for hybrid,hierarchical systems. This project will develop techniques for computational modeling andanalysis, and a computational infrastructure to support the autonomousmodel-enabled control of multiple hybrid systems. Model-enabledcontrol is realized by embedding rich computational device models andassociated reasoning and analysis machinery directly into on-linecontrol systems. Such model-enabled controllers can autonomously adaptto new high-level task requirements and unanticipated changes in theirenvironment. They also can detect deterioration or failure of devicesand compensate for such contingencies through reconfiguration of theremaining components. A multi-level model-enabled control architecturecan provide the machinery for coordination of tasks among multipleautonomous systems. The autonomous control of multiple airborne vehicles for space andearth science, surveillance, and weather mapping is an example of acomplex problem for which traditional simulation and controltechniques are inadequate. Each vehicle in the group or fleet is, ofitself, a complex hybrid dynamic system. Traditional approaches tothe fleet control problem require human operators who determinemission-level tasks, convert the tasks to detailed commands, andcontinually upload the commands to individual vehicles. The result isinflexible, custom-built, real-time operations with very littleability to adapt and react to unexpected situations caused by failuresin subsystems, changes to the vehicle environment, and theintroduction of new goals and tasks. Model-enabled control isparticularly suited to the task of autonomous control of multipleairborne vehicles. We will, therefore, use that problem domain to testand demonstrate the technology developed in the project. Developing software in support of model-enabled control presentsdiverse computational challenges. The success of our research thusrequires a coupling of expertise from multiple disciplines of scienceand engineering including control theory, artificial intelligence,model-based reasoning, and hybrid systems modeling and control, aswell as disciplinary expertise in the design, modeling, and control ofairborne vehicles. This project is proposed by a multi-disciplinaryteam of researchers from the Stanford Computer Science DepartmentísKnowledge Systems Laboratory (KSL), the Stanford MechanicalEngineering Departmentís Center for Design Research (CDR),Stanfordís Aeronautics & Astronautics Department (Aero/Astro),Vanderbiltís Computer Science Department, and an industrialcollaboration with the Systems and Practices Laboratory (SPL) at theXerox Palo Alto Research Center. The proposed approach is to build on the previous work of the teammembers. Specifically: The modeling paradigm is based on compositional and hybrid modeling techniques pioneered at Stanford KSL and Xerox SPL intermixed with engineering and control theory modeling techniques from Aero/Astro and mechanical engineering; Model-based reasoning about structure and function has been investigated for many years at Stanford KSL and provides a core for the model-enabled aspects of the control architecture;The hybrid control architecture builds on hybrid modeling and simulation, mode identification, and diagnosis methods for complex physical systems developed at Vanderbilt; A new systems architecture for fleets of airborne vehicles has been developed at Stanford Aero/Astro; andThe use of computer-interpretable design rationale in controller design has been explored at Stanford CDR. Our research will bring together these independent strands of researchto produce systems with dramatically enhanced capabilities inautonomous coordinated control. The results will have impact on thosewho design and develop controllers, especially in control of airbornevehicles, and on those who simulate, diagnose, and verify hybridsystems. The research will result in new computational capabilitiesand tools for both computer scientists, control engineers, and designengineers. s, control engineers, and designengineers.
|
