1 Copyright © 2006 by ASME Proceedings of IMECE06 2006 ASME International Mechanical Engineering Congress and Exposition November 5-10, 2006, Chicago, Illinois IMECE2006-15302 THE VEHICLE AUTOPILOT: SIMULATANEOUS ROBUST CONTROL THROUGH PARAMETRIC ADAPTATION Haftay Hailu Graduate Student Sean Brennan Assistant Professor 318 Leonhard Building University Park, PA 16802 Phone: 814-863-2430 Fax: 814-865-9693 e-mail: sbrennan@psu.edu ABSTRACT This work considers the problem of robustly controlling systems that have an implicit parametric coupling, and specifically considers the problem of lateral control of passenger vehicles at highway speeds. Passenger vehicles collectively have a wide range in dynamic behaviors mainly due to the ranges in size between different models. However, as vehicle size increases, the length, mass and mass moments of inertia also increase in predictable relationships that strongly couple these parameters to each other. The proposed control technique exploits this inherent parametric coupling in order to design a single robust controller that can be easily adapted parametrically from vehicle to vehicle. Parameter decoupling in the design model is achieved in the control synthesis step using a dimensional transformation. The resulting design model presents a system representation suitable for robust control of a very wide range of passenger vehicles using only a dimensional rescaling. This method is distinguished from prior work in that the structure of parametric dependence is included in the controller synthesis. The resulting design is tested on a scaled vehicle test setup developed at Pennsylvania State University. Both simulation and experimental results have shown the effectiveness of the technique for the proposed application. 1. INTRODUCTION This work discusses a robust, simultaneous control technique for systems whose system parameters are inherently coupled. Human- or naturally-optimized systems will likely exhibit a property where many of the system parameters entering the dynamic model are strongly interrelated. This arises because the key dynamic parameters of a system are generally the same parameters that must be optimized to satisfy design criteria in the system build. A physical example of a collection of systems whose behavior is similar yet scaled along key dynamic parameters is the family of passenger vehicles. For example: a passenger vehicle larger than average tends to be longer, heavier, and with a larger mass moment of inertia than average as well. Additional generalizations can be made between vehicle size and the tire force generation performance, the suspension behavior, etc. These relationships between length, mass, inertia, etc. obviously do not follow an exact functional relationship. But if one simply knows that the system under consideration is a modern production passenger vehicle, one can infer general estimates of many parameters if given just one parameter, mass for instance. This inference can be formalized as equations describing coupling parameter relationships. The application of a generalized robust control and/or guidance technique in automotive applications is not as extensive as in the aerospace industry, at least as reported in public literature. However, robust control implementation are gaining increased interest in applications of Automated Highway Systems (AHS) [1, 2]. A robust ∞ H loop-shaping controller was designed in [1] and a nonlinear robust controller was developed for lateral control of heavy trucks in automated highways in [2]. In most vehicle models, the vehicle velocity appears as a free parameter due to the significant changes in the vehicle dynamic model as a function of velocity, changes that sometimes change an open-loop stable model to an unstable model with increasing speeds. Thus, gain-scheduling is often required and used. To address this velocity dependence, a gain- scheduling controller was designed in [3] and an LPV controller in [4]. Additional application are described in [5-9]. While scaling theory is an old subject and has been applied to dynamical and structural systems analysis, its application to control of these same systems is very limited and has been seen in literature only during the last decade. One of the most recent and well developed work in this area is the works of Brennan and Alleyne [7, 10, 11]. Previous work by Brennan [7] have shown the advantages of using the dimensionless representation in vehicles for robust control design. Specially, Brennan [7] has shown the achievement of tight frequency-domain variations