Decoupled design of robust controllers for nonlinear systems: As motivated by and applied to coordinated throttle and brake control for automated highways

This dissertation examines the longitudinal control task for automated vehicles as an application of, and a paradigm for, the development of general results in robust nonlinear control. While the inherent nonlinearities and uncertainties of vehicle dynamics make longitudinal control--modulating brake and throttle inputs to achieve a desired speed or spacing--a natural choice for such techniques, several factors preclude straightforward analysis. From a physical standpoint, the need to switch between the dissimilar dynamics of the engine and brake systems presents a serious obstacle to analysis. The difficulty is compounded when the limited control authority of the engine and the rate limits exhibited by both the engine and brake dynamics upon release are also considered. From a conceptual standpoint, automated highways are inherently a hierarchical structure. Hence, controller design that assumes a single designer from actuator to output (the paradigm for feedback linearization, among other techniques) seems fundamentally out of place in this environment.; To help establish the physical characteristics of the vehicle, the dissertation begins by presenting a model of brake dynamics suitable for simulation or control. In its entirety, this structure consists of three primary elements: an "upper" sliding controller defined by the vehicle control task, a switching logic (with hysteresis) for determining whether throttle or brakes are required and the individual "lower" surface controllers for the brakes and the engine.; The theoretical heart of this work involves a more rigorous treatment of this decoupled design process for a general class of nonlinear, SISO systems characterized by a series of nested control loops. In this design, the sliding surface structure is augmented by a series of filters to produce a Dynamic Surface Controller (DSC). This structure removes questions of numerical differencing and applicability to systems of order n that have plagued other multiple-surface sliding controller designs. While motivated by the automated highway application, these results are still quite general. With some additional mathematical treatment of rate limits and switching, however, the specifics of the coordinated throttle and brake coatroller can be analyzed in a similar manner. This is demonstrated in detail for the case of speed control by determining tracking bounds for the closed loop system in the presence of model uncertainty, rate limits and the hysteresis element in the brake/throttle switching logic. Experimental results exhibit performance even better than that guaranteed, with errors of approximately 0.1 m/s while tracking a combined acceleration and deceleration profile. Experiments with spacing control show that spacing errors of only 20 centimeters are produced while attempting to maintain a constant spacing of 2 meters behind a maneuvering lead vehicle traveling at highway speeds. (Abstract shortened by UMI.)...