Autonomous driving represents a technological leap forward that will likely solve key aspects of the transport problem and so have beneficial effects for some ciritical social and ecological issues as well. However, testing and validation of the Highly Automated Vehicle (HAV), so as to guarantee safety, is one of the most challenging tasks that still prevent the HAV from commercial release. As HAVs will include many functions which can be updated frequently and require re-evaluation in short time, fast re-evaluations will be needed to prevent new dangers arising from the updates. This research work is focused on developing such a methodology. The key idea is to replace on-road testing ? well defined testing method for classical vehicles ? by accident statistics and to use model based methods as well as Design of Experiments to determine a limited set of cases to be tested.