Collision avoidance system for fixed obstacles-fuzzy controller network for robot driving of an autonomous vehicle

Abstract

A Collision Avoidance System (CAS), which overrules the driver in a critical situation, by steering and/or braking has to be better and more reliable than the driver himself. The driving maneuver is complex and difficult to calculate by traditional mathematical models. Therefore, an ACC car with extended sensors for object detection and a human driver were taken in order to get the data how the driver avoids the Collision with a fixed object in the driving lane. Afterwards, this data was used in order to develop a fuzzy controller network of full collision avoidance for fixed objects. The effectiveness and the robustness of the more than 300 rules of the Fuzzy Controller Network were tested by using the same ACC car, but driven by a robot on the driver seat. The result of these tests are presented in this paper.