Safety Testing for Autonomous Systems in Simulation

Safety is a critical concern in unmanned autonomous systems. Simulation environments that use a virtual environment to provide realistic inputs with the autonomy algorithms in the loop have proven valuable for replicating closed-loop behavior seen in live testing. These simulation environments can also supply the varied conditions (unusual weather, moonlit nights) required for safety testing that are often unachievable in all but limited instances in the real world. JRM Technologies proposes to design and develop a Robotic and Unmanned System Safety Testing Environment (ruSafe) to add to the TARDEC RIDE Simulation Framework. This innovative environment will add key new functionality to the existing simulation infrastructure and provide an automated safety testing and analysis harness: (1) Unattended testing capabilities; (2) Support for statistical techniques such as Monte Carlo; (3) Data recording and management; and (4) Automated analysis techniques for identification of failure modes and display of results. By starting with an established simulation infrastructure, the efforts can focus immediately on the critical new functionality outlined above. Enhancing a proven simulation environment will allow an early assessment of the new functionality and minimize risk.