Computational Intelligence Techniques for Car Crash Detector Optimization

Modern occupant restraint systems rely on primary (seat belt) and secondary restraint (airbag) activities. Different types and different intensities of car crashes require different reactions of the restraint system within time spans down to ten milliseconds. Thus, the state signals generated by acceleration sensors have to be processed fast and accurately. We apply evolutionary methods to fine-tune decision trees constructed using car crash (training) data, and to select a subset of sensor features yielding best performance of various classifiers including artificial neural networks.

Master Student


Financial Support

Project Duration