engleski

Time Optimal Control of Piecewise Affine Systems

This dissertation investigates the constrained time optimal control of nonlinear processes modeled in a discrete-time piecewise affine form. The control system states are in minimum time steered into the computed invariant set in which it is guaranteed that they can be kept for an arbitrarily long time. The control system design is based on computations with polytopic predecessor sets. We introduce procedures for efficient computation of the maximum controlled invariant set and of the time-optimal control law which reduces the number of switches between model dynamics. For the case of the polytope-bounded model uncertainty we introduce a new algorithm for the computation of robust predecessor sets and consequently robust invariant sets. The corresponding computation time is for complex models reduced by several orders of magnitude. We investigate the clustering-based identification procedure for piecewise affine process models. The original procedure is modified in order to enhance the approximative properties of the identified model. The developed identification and control methodology is validated on two challenging industrial control problems: the electronic throttle servosystem and the torque control of switched reluctance motors. Experimental results obtained on the electronic throttle servosystem show fast and accurate transients which is very important for the car driveability and for reducing the fuel consumption and the polluting gas emissions. The use of switched reluctance motors is significantly limited in the variable-speed applications due to the high level of induced motor torque ripples. We show how the torque control law design based on polytopic invariant sets can be applied to limit the torque ripples to their physical minimum.

constrained time optimal control; invariant set; polytope; discrete-time piecewise affine model; robust control; piecewise ARX model identification; clustering; automotive electronic throttle; switched reluctance motor

nije evidentirano