| This thesis investigates the problem of regulation-level vehicle control for the automated highway system (AHS). The issue of longitudinal and lateral coupling, as it affects tracking performance and robust stability, is explored in detail. To begin, a 26-state simulation model is proposed. Drawing from a number of earlier modeling results, this model offers the necessary level of fidelity for the study of combined vehicle maneuvers in typical freeway operations, while minimizing the level of complexity. Specifically, it includes descriptions of the vehicle sprung mass, engine, brakes, transmission and drive-train, suspension, and tire behaviors. Further, two simplified control models are also presented that are more amenable to analysis and control design.; Utilizing these simplified models, it is shown that the coupling between the lateral and longitudinal dynamics can become appreciable during severe highway maneuvers, such as those encountered during lane changes, emergency obstacle avoidance, or in inclement weather and road conditions. To corroborate this result, various coupled and de-coupled control strategies are used in vehicle simulations over a wide range of maneuvering profiles and operating conditions. The closed-loop results indicate that tracking performance does improve under the coupled control strategy, and that the improvement becomes more pronounced as the maneuvers become more demanding.; In order to facilitate coupled control design for multiple input, multiple output (MIMO) systems with unseparable input nonlinearities, sufficient conditions are presented for the existence of an approximate (numerically obtained) control which guarantees bounded tracking.; The role of previewed road curvature information in improving lateral tracking performance is also examined. However, due to the abrupt nature of lane changes in US highways, a straightforward application of the previewed information is not desirable, for it leads to excessive steering demands that may result in overshoot or may violate passenger comfort demands. Consequently, a virtual trajectory is computed from the previewed curvature and is referenced to the road centerline. Moreover, this trajectory is optimal in the sense of minimum jerk, minimum error (between virtual trajectory and road centerline).; In addition to controller design for individual agents, the issue of vehicle stability within an interconnected system is further investigated. It is known that individually stable constituents may form unstable systems due to the dynamic interaction among the members. Sufficient conditions for the stability of the interconnected system, string stability, are also known. This thesis proposes an analysis strategy and equivalent sufficient string stability conditions for sampled-data controlled systems. The tools are applied to prove the stability ot the proposed hybrid longitudinal vehicle-following, point-following and hybrid lateral heading angle, lane-following algorithms.; The hybrid longitudinal scheme is superior to the constant-spacing vehicle-following strategy since it: provides absolute ground-truth referencing; places much less demand on the radios; and is more robust, in string stable sense, to velocity estimation errors. This scheme is also superior to the point-following strategy because vehicles can be much more safely close-packed, i.e. in a platoon, due to its incorporating the dynamics of the preceding vehicle in the control.; By the same token, the proposed hybrid lateral strategy is superior to both exclusive vehicle-following and lane-following algorithms because it draws from the advantages of both to buttress their respective weaknesses. (Abstract shortened by UMI.)... |
