An integrated approach to the design of linear dynamic network-based systems

This research focused on the development of an integrated approach to the design of linear dynamic network-based control systems that incorporate biologically inspired concepts. Hierarchical networks were constructed to coordinate the behavioral strategies of goal-seeking elements to produce a control system able to reconfigure a two-link manipulator. Homeostatic and heterostatic goal-seeking behavior was modeled using noncooperative dynamic game theory and quadratic cost functions that were constructed by the network elements in response to the connective environment established by the system designer. The operation of the network-based control systems was consistent with the biological notions of goal-seeking behavior, specialized network structure, pretuned network connectivity, prerotation of the network state vector, and command anticipation.; An alternative approach to network design, based on the maintenance of global optimality, was developed as a means for interconnecting goal-seeking elements that pursue quadratic-tracking goals and which do not modify their behavioral strategies in response to structural perturbations. A coupling factorization condition and an equality constraint significantly reduce the number of design-related degrees-of-freedom and guarantee the preservation of optimality and closed-loop stability in the presence of nonzero connectivity.; Qualitative and quantitative methods of analysis were used to investigate several features of the control systems. Robustness was evaluated using an estimated rate of performance degradation under perturbed task conditions. Principal component analysis provided information on the number of dominant orthogonal directions of time-varying connectivity and interaction-exchange among the states of the system. Finally, graph theoretic analysis measured the fragmentation, reachability, and hierarchy of the control systems.; The design procedure was successfully applied to four reconfiguration tasks. Control systems created with homeostatic goal-seeking elements were more robust than those constructed with heterostatic elements and unique properties characterized the performance and dynamic structure of each type of control system. It is concluded that the integrated approach to dynamic network design is valid and that auxiliary conditions can be used to simplify development.