A driver monitoring and warning system called DAISY (Driver AssIsting SYstem) is presented, which adapts its warning messages to warning thresholds acceptable to the driver. This is achieved by the use of a model of the individual driving behaviour of the driver actually driving. Artificial neural networks based on the ART architectures were used for on-line learning to realize this model. Besides a brief survey of the main features of DAISY this paper emphasizes the modelling of the driving behaviour with ART networks. The way the ART networks are adpated to the problem is described. This includes the realization of a hierarchical ART structure and of a situation specific weighting of the components of the feature vector. The network parameters were adjusted through genetic algorithm optimization. (Author/publisher).
Abstract