Abstract
A methodology and toolset is introduced and used for assessing the performance of safety-critical automotive vehicle-control system software. The method is based on three elements: (i) formal system analysis and decomposition, (ii) directed Monte Carlo testing, and (iii) statistical analysis of the results. By applying this method to an experimental fully automated cooperative system for lateral vehicle control, an estimate of system performance is obtained under a wide variety of conditions. The test strategy used was found to be effective for finding anomalous or incorrect system behavior. (A)