A hybrid system approach to a class of intelligent control systems

In intelligent control of complex systems, one is faced with the problem of providing stability, performance, and robustness at the closed-loop real-time level, as well as decision-making control with guaranteed performance at the supervision level. Unfortunately, it is difficult to formally derive a finite state representation of a closed-loop continuous-state dynamical system that is suitable for upper-level decision-making functions with closed-loop stability of the entire system. Any discrete-event representation of dynamical systems must encapsulate the relevant behaviors and events needed for effective decision-making control. A Hybrid System Framework is needed for considering simultaneously the real-time control and decision-making issues.; Hybrid systems have emerged as a technique for modeling and analyzing a class of autonomous control systems containing both continuous-state dynamics and logical clauses. We consider a formulation of hybrid systems where a continuous-state plant is supervised by a discrete-event controller. The discrete-event controller activates a closed-loop behavior of the supervised plant by choosing a controller, a reference input, and a sensor output from a library of real-time controllers, reference trajectories and output functions, respectively.; We address the hybrid controller synthesis problem, that is designing a discrete-event supervisor that coordinates low-level continuous-state control policies in a stable and safe manner. Conventional Lyapunov theory cannot be applied to hybrid systems directly. We study the stability of a class of hybrid dynamical systems by means of non-smooth Lyapunov functions. Finally, a hybrid controller has been successfully implemented to swing up and balance an underactuated mechanical system.; Nonholonomic mobile robots have been used by many researchers as a testbed of intelligent control architectures. Unfortunately, conventional feedback linearization techniques cannot be applied to nonholonomic systems directly. We develop a control structure that makes possible the integration of a kinematic controller and a neural network computed-torque controller for nonholonomic mobile robots. A combined kinematic/torque control law is developed using backstepping and stability is guaranteed by Lyapunov theory. The neural network controller proposed in this work can deal with bounded disturbances and/or unstructured unmodeled dynamics in the vehicle.