1 Neural Networks for Automated Vehicle

Scope and Purpose. Inspired by the findings about neurons, a n d especially by some effective training procedures found in the late 70's, neural networks have seen a resurgence in the research community. I t s philosophy assumes simple but massive interconnected processing uni t s as the basic structure for information processing. Problems that a r e difficult to tackle with conventional symbolic-based techniques, seem t o be more readily solvable with a neural network approach (e.g. perceptual tasks, associative memory). This paper focuses on such a problem: t h e dispatching of vehicles and crews. This kind of problem is found, for example, in the context of parcel pick-up and delivery services or pr iori ty mail. The task of assigning a particular vehicle or driver to a new reques t is done by a human dispatcher, and his(her) decision process typically involves a lot of experience, judgment and expertise. We thus describe i n this paper a neural network model that can learn to reproduce t h e decision process of an expert dispatcher by using examples of his(her) previous decisions.