Validation of decision-making in artificial intelligence-based autonomous vehicles

ABSTRACT The autonomous vehicle technology is considered as a significant market disruptor for multiple industries. However, to reach this potential and to be accepted by the public, autonomous vehicles must be proven to be reliable and safe. Therefore, validation is essential for improving the public trust for autonomous vehicles and deploying them for everyday transportation activities. There have been a number of significant efforts on validation of autonomous vehicles; and real-life testing and test tracks have been the major platforms for these activities. However, simulation has also been gaining popularity due to its advantages in cost, time and safety. In this paper, we present a simulation scenario generation methodology with pseudo-random test generation to validate the decision-making system of autonomous vehicles. The methodology separates the validation concerns and focuses on generating scenarios that test the decisions taken by the vehicle. The implementation demonstrates the capabilities and the efficiency of the approach.