Autonomous behaviors for interactive vehicle animations

Abstract

We present a method for synthesizing animations of autonomous space, water, and land-based vehicles in games or other interactive simulations. Controlling the motion of such vehicles to achieve a desirable behavior is difficult due to the constraints imposed by the system dynamics. We combine real-time path planning and a simplified physics model to automatically compute control actions to drive a vehicle from an input state to desirable output states based on a behavior cost function. Both offline trajectory preprocessing and online search are used to build an animation framework suitable for interactive vehicle simulations. We demonstrate synthesized animations of spacecraft performing a variety of autonomous behaviors, including <i>Seek, Pursue, Avoid, Avoid Collision</i>, and <i>Flee</i>. We also explore several enhancements to the basic planning algorithm and examine the resulting tradeoffs in runtime performance and quality of the generated motion.