Autonomy Artificial Intelligence Robotics (AAIR)


by Mads R. Jeppesen; Sigurd A. Madsen; Ole Balling


Development and assessment of autonomous vehicle capability are relying on simulation software for time and cost efficiency. The value of such simulations are significantly dependent on minimizing the gap from simulation to real environment performance of systems. The simulations for off-road autonomous vehicle assessment are in particular challenging due to the complex nature of natural terrains and their virtual representations, vehicle-terrain interactions during soft soil maneuvering, and the integration of sensors and their output in virtual generated terrains. This paper presents the early development of a software tool aimed at simulating custom autonomous off-road scenarios generated from their real world counterparts. The effort is an important step in generating confidence in simulation based testing of autonomous systems as a forerunner for purely virtual generated scenarios for autonomous systems evaluation.