Guide to Anticipatory Collision Avoidance

摘要

This chapter provides a step-by-step explanation of a modern anticipatory collision-avoidance algorithm and discusses how to optimize its implementation. The two most important forces on an agent's path is a driving force, which pushes an agent to its goal, and a collision avoiding force, which resolves collision with neighboring agents in an anticipatory fashion. In predictive avoidance method, some perturbation is introduced in the collision-avoidance routine to account for the uncertainty that an agent has in sensing the velocities and radii of its nearby neighbors. The optimal reciprocal collision avoidance (ORCA) method works by defining a set of formal constraints on an agent's velocity. One exciting area of interest has been applying anticipatory collision-avoidance techniques to robots. Other interesting challenges include incorporating an agent's anticipation into its character animation or adapting techniques such as ORCA to account for stress, cooperation, and other social factors.