Many real world problems quirea degree of flexibility that is to achieve using hand algorithms. One such domain is vision-based autonomous driving. In this task, the dual challenges of a constantly changing environment coupled with a real processing constrain the flexibility and of a machine learning system essential. This describes just such a learning system, called (Autonomous Land Vehicle In a Neural Network). It presents the neural network architecture and training techniques that allow to drive in a variety of including singlelane paved and unpaved roads, multilane lined and unlined roads, and obstacle-ridden on- and road environments, at speeds of up to 55 miles hour.

Many real world problems requireadegree of #exibility that is di#- cult to achieve using hand programmed algorithms. One such domain is vision-based autonomous driving. In this task, the dual challenges of a constantly changing environment coupled with a real time processing constrain make the #exibility and e#ciency of a machine learning system essential.

Abstract. Driving is a multitasking activity that requires drivers to manage their attention between various driving- and non-driving-related tasks. By modeling drivers as continuous controllers, the discrete nature of drivers ’ control actions is lost and with it an important component for characterizing behavioral variability. We propose the use of cognitive architectures for developing models of driver behavior that integrate cognitive and perceptual-motor processes in a serial model of task and attention management. A cognitive architecture is a computational framework that incorporates built-in, well-tested parameters and constraints on cognitive and perceptual-motor processes. All driver models implemented in a cognitive architecture necessarily inherit these parameters and constraints, resulting in more predictive and psychologically plausible models than those that do not characterize driving as a multi-tasking activity. We demonstrate these benefits with a driver model developed in the ACT-R cognitive architecture. The model is validated by comparing its behavior to that of human drivers navigating a four-lane highway with traffic in our fixed-based driving simulator. Results show that the model successfully predicts aspects of both lower-level control, such as steering and eye movements during lane changes, as well as higher-level cognitive tasks, such as task management and decision making. Many of these predictions are not

The Illinois Roadway Simulator (IRS) is a novel, mechatronic, scaled testbed used to study vehicle dynamics and controls. An overview of this system is presented, and individual hardware issues are addressed. System modeling results on the vehicles and hardware are introduced, and comparisons of the resulting dynamics are made with full-sized vehicles. To address the realism factor of using scaled vehicles, comparisons are made between dynamic responses of full-scale and IRS-scale vehicles. The method of dynamic similitude is a key to gaining confidence in the scaled testbed as an accurate representation of actual vehicles to a first approximation. The IRS is then used in a vehicle control case study to demonstrate the potential benefits of scaled investigations. The idea of driver-assisted control is formulated as a yaw-rate model-following problem based on the representation of the driver as a known disturbance model. The controller is designed and implemented to show that the vehicle's dynamics can be changed to match a prescribed reference model.

INTRODUCTION As the use of control systems in vehicles has increased the past several yem-s.!the safety', per/'ormance, and reliability of vehicles has improved, primarily duc to the implementation of automated controllers. A significant amount of research has been done on vehicle control (Tomixuka and Hedrick, 1995, Shlado,/er, 1995), but much of this work has been limited to simulation because the use of a lull sze vehicle to test controllers is often prohibitively expcnsiw' as well as dangerous. The focus of this research has bccn to develop a scale version of a vehicle and a roadway for safe and economic testing of these types of controllers. To that en& the Illinois Roadway Simulator (IRS) has been developed. The IRS is m experimental testbed consisting of scaled vehicles runningl on a simulated road surface. There arc several advantages of the IRS over f11 scale vehicle testing. Firsk the availability of scale vehicle components makes construction sitrple and very cheap. The du