Verification of Safety for Autonomous Unmanned Ground Vehicles

The existing tools for hardware and software reliability and safety engineering do not supply sufficient solutions regarding Al (Artificial Intelligent) adaptive and learning algorithms, which are being used in autonomous robotics and massively rely on designer experience and include methods such as Heuristic, Rules based decision, Fuzzy Logic, Neural Networks, and Genetic Algorithms, Bayes Networks, etc. Since it is obvious that only this kind of algorithms can deal with the complexity and the uncertainty of the real world environment, suitable safety validation methodology is required. In this paper we present the limitation of the existing reliability and safety engineering tools in dealing with autonomous systems and propose a novel methodology based on statistical testing in simulated environment.