How would you validate that a simulation-generated dataset improves performance on rare obstacles for Autonomous Solutions?
