Visual perception is a major component in the driving task. Learning how to drive will be determined, at least in part, by learning how to perceive the traffic environment and how to reason with time - spational relationships. The Prospect project aims at enhancing driver's education. This includes teaching visual strategies. For this purpose a cognitive analysis has been carried out and a simulation of the driver is realized in a production rule system. The analysis consists of three phases. Firstly, a taxonomy of visual tasks has been established by running a normative analysis on the driving task. Next, a algorithmic model has been formulated which is capable of describing all tasks in the taxonomy. This model is inclusive of definition of layers of internal representation, elementary internal processes on and the information flow between and within layers of representation. Lastly, the algorithmic model is implemented in a production rule system, Soar, in order to test the model's functionality. (A)
Abstract