Driving environment recognition for adaptive automotive systems

With the rapid development of electronics and the growing demand for higher performance with respect to safety, driveability, fuel efficiency, and emissions, modern automotive systems are required to perform increasingly sophisticated tasks. To meet these challenges single type controls for each subsystem will tend to be integrated by an overall intelligent control system which is able to perceive the present situation and adjust adaptive vehicular components accordingly. To take a crucial step towards intelligent automotive systems the problem of environment recognition is addressed and a neuro-fuzzy approach for the identification of the driving situation based on available sensor information is introduced. It uses fuzzy logic for the classification task, generated and optimized by means of a neural network, and allows the bidirectional conversion between the fuzzy and neural domain. The proposed method leads to superior classification results and reduced development time compared to “manual” system design