Distributed Collision Detection and Resolution

Multiplayer, online computer games often distribute game-object state to client machines in order to improve game scalability and responsiveness. Network latency and jitter are concerns in this context, although the impact is reduced by the use of predictive techniques such as dead reckoning. These techniques, however, introduce consistency concerns for important and hard to predict behaviours, such as object collisions. In these cases a centralized authority or client/server architecture is typically used to ensure strong consistency, limiting game scalability. In this work we propose a motion-lock protocol for distributed game collision detection and resolution. The motion-lock protocol improves performance of motion prediction by giving stations time to communicate and agree on the detected collisions. This reduces the divergence of object states and post-collision trajectories. Offline and online simulation results show the motion-lock protocol is able to maintain strong consistency in collision count and reduces post-collision deviation with a small sacrifice of 3-4% in responsiveness of player controls. Qualitatively, the visual result of the collision response is greatly improved. With the motion-lock protocol, multiplayer online games can offload basic collision detection and resolution to game clients, increasing scalability without overly sacrificing consistency.