Reciprocal Velocity Obstacle (RVO) is a velocity-based collision avoidance method used in microscopic crowd and multi-agent simulation, wherein each agent selects a new velocity outside the set of velocities that would lead to a collision with any other agent within a given time horizon, with the "reciprocal" principle ensuring that responsibility for avoidance is shared symmetrically between pairs of agents to prevent oscillatory or conflicting maneuvers. It matters because it provides a computationally efficient, locally computable, and mathematically principled framework for generating smooth, collision-free motion in dense crowds without requiring centralized coordination, making it well-suited for real-time simulation of large pedestrian populations. Key variants include the original Velocity Obstacle (VO) formulation, Hybrid Reciprocal Velocity Obstacles (HRVO), and Generalized Reciprocal Velocity Obstacles (GRVO), which extend the approach to handle non-holonomic constraints, larger agent geometries, and heterogeneous groups with differing speeds and sizes.
Dictionary term