CAREER: Coordinated Control from Deformable Virtual Structures with Dynamic Communication

University Of Washington, Seattle WA

Investigators

Abstract

The use of multiple vehicles for the achievement of a joint task produces an operational setting with greater flexibility, robustness and coverage for a lower cost than is possible with single vehicles or stationary sensing networks. When the operational environment for such vehicles is complex, dynamic or contains operational hazards, individual vehicles must have the ability to respond quickly and with a high level of maneuverability. The focus of the work in this project is the development and implementation of control theoretic methods for coordination of groups of agile vehicles coupled through communication and operating in dynamic environments. The need for development of these methods arises when traditional design methodologies fail due to the operation of autonomous systems in the presence of a strong coupling between dynamic information flow and underlying system dynamics. The work will be primarily driven by challenges arising in the coordinated control of agile underwater vehicles, however, the results will also have high impact for tasks such as vision-based reconnaissance with aircraft in hazardous urban environments. The particular combination of agility, dynamic communication and operational uncertainty requires that the communication and coordination elements of coordinated control be considered jointly rather than independently. The objectives of this project are (1) develop a deformable virtual structures approach to cooperative control which simultaneously captures the constraints imposed by limited communication and by vehicle constraints, and (2) develop dynamic communication structures that are sufficient to guarantee stability and coherence of the virtual structure. The group of vehicles is treated as a virtual, deformable body which is effectively underactuated due to velocity and motion constraints on the constituent vehicles and the physical constraints necessary to maintain communication between the vehicles. The constituent vehicles then become shape actuators for the evolution and motion planning for this new class of virtual mechanical systems with lumped, dynamic mass properties. The communication structure of the vehicles is assumed to be dynamic with the transmission connections treated as control variables. Construction of communication patterns is addressed via periodic sequences chosen using geometric and graph theoretic tools. The governing equations for the resulting virtual structure then describe an underactuated mechanical system with both discrete time actuation and oscillatory dynamics. Analysis and construction of feedback control for such systems is addressed by developing a new notion of averaging and state feedback. Research targeted to applications and implementation in underwater autonomous vehicles provides a captivating venue for education at all levels. The issues arising in such settings have broad theoretical implications as well as the appeal of hands-on experience. To strengthen and broaden interaction between control theory and fields such as biology and computer science, undergraduate courses in control theory will be redesigned to make the tools of control theory material accessible to students with non-traditional engineering backgrounds. Additional graduate courses will also be developed that provide not only a rigorous theoretical background in individual elements of control theory or communications but that also provide tools for developing interdisciplinary methods and application of those methods. Closely integrated with the research and educational components of the project are mentoring and outreach activities based around the construction of an interactive web site. This component involves not only undergraduate and graduate students as developers, but K-12 students and educators as active users and participants in the research lab. The site will contain scientific information about the different components of autonomous vehicle design and control as well as interactive tools for exploring application of control and communication methodologies to lab vehicles.