This paper proposes a mathematical norm against which driving simulation results (one driver and one vehicle interacting with an environment of fixed and moving obstacles) can be directly compared. The norm is, in effect, an optimal control strategy (in terms of steering, braking and accelerating) given specific constraints on driver's visual span, dynamic equations of the vehicle, and the cost function or trading relation between penalty for colliding with an obstacle and the penalty function of increased control effort or time. The mathematical model, embodied in a digital computer, is being used by the authors in conjunction with both a closed-circuit TV laboratory simulator and actual road tests. To illustrate the idea, some empirical trajectories obtained on the driving simulator are presented together with various computed trajectories which are optimal for particular cost functions and a simplified dynamic model of the automobile. Three appendixes describe the simulator, give comparable data from both simulator runs and road tests with a standard automobile, and provide a simple numerical example of dynamic programming. /author/.
Abstract