Test and Evaluation of Autonomous Vehicles Using Simulated Environments

Test and evaluation (T&E) of autonomous vehicles presents a challenge as the vehicles may have emergent behavior and it is frequently difficult/impossible to ascertain the reason for software decisions. Requiring the need for T&E during the complete lifecycle of a vehicle’s development and fielding, requiring T&E after deployment due to emergent behavior and software updates. A software framework has been developed to support the development of autonomous software such that it can be tested during the complete lifecycle, migrating through the virtuality/reality spectrum. The paper demonstrates the application of simulated/virtual reality/augmented reality/physical environments to test and evaluate the sense, plan, and act phases of an autonomous system to ensure proper operation with little to no physical activity from the system allowing for safe testing of the system without endangering either the system or the environment. The demonstrations consist of the processes of constructing models for simulating the physical world through various sensors to transpose onto the autonomous software from a virtual environment. The autonomous software cannot distinguish between physical and simulated data; allowing for more fidelity in the testing of the autonomous software in a simulation environment. Then a process for various levels of simulation can be introduced, such that, the testing of the software goes from a fully virtual environment to a fully physical environment ensuring proficient results without compromising safety. Development of VR/AR testing is demonstrated using an autonomous rover at Old Dominion University\u27s Autonomous Modeling and Simulation Engineering Lab