Publication

The report introduces a general method for verification of Collision Avoidance Systems. The method is based on a principle called mileage multiplying, consisting of nine steps. By using this method it might be possible to significantly reduce the need for data collection associated with the system verification. Targets, i.e. vehicles and pedestrians, are identified with aid from a radar and a camera. Information from these sensors are fused together by a sensor fusion algorithm. The system then determines if an intervention should be triggered. It is relatively easy to verify that a Collision Avoidance System does an intervention when it should, but it is a much more complex task to conclude that the system will not make any so called unnecessary interventions. These unnecessary interventions might be annoying for the driver and should always be avoided. The mileage multiplier method consists of detecting interesting radar targets by analyzing collected traffic data. The radar targets are then modeled and inserted into the traffic data by using different statistical approaches. By simulating the Collision Avoidance Systems with the modified data as input, it is possible to analyze the system behavior to possibly find a measure of robustness for the specific situation. Data have been generated in a random fashion to better illustrate the strengths of the general method. The method should be used as a complement to conventional system verification. If a weakness of the Collision Avoidance System should be found, it is recommended to perform real life trials in order to validate the results.