ABSTRACT
The reciprocal velocity obstacles (RVO) algorithm and its descendants, such as hybrid reciprocal velocity obstacles (HRVO) and optimal reciprocal collision avoidance (ORCA), have become the standard for collision avoidance in video games. Much of the popularity of velocity obstacles (VO) methods can be traced to the ease of understanding them as prohibited shapes in velocity space. ORCA's cornering problem is one example of a larger class of problems exhibited by VO methods when the assumption of constant preferred velocity is violated. The RVO algorithm softens VO constraints, treating projected collisions as only one aspect when “scoring” a velocity candidate, making it a utility method. VO methods can adapt to occasional changes in heading without much issue. Original VO methods treated the duty to avoid collisions as absolute. Making progress in their preferred directions was a secondary concern, only coming into play when choosing among a set of noncolliding and therefore valid trajectories.
