Two driver models based on neural networks, for speed and acceleration control as well as for lateral control will be introduced. In contrast to conventional models, driving is not represented as control with respect to given input functions, but as a reaction to external situations. The first model for speed and acceleration control is divided into two parts: a neural network classifies the street geometry as a variable which can be interpreted as the representation of action. These are used by the second part, the fuzzy controller, which adjusts speed and acceleration. In order to simulate a double lane change, a controller based on back propagation networks will be introduced in the second model. (A)
Samenvatting