AI Level of Detail in Multiplayer Games

Real-time multiplayer games come with a strong set of constraints. The device that runs the simulation needs to handle the complexity of dealing with several players: replication, prediction and reconciliation, anti-cheat, etc. When the player machine hosts the game, client tasks such as graphics, audio, and UI also need to run and will take up most of the system resources. With dedicated servers, the system specs are often kept to the minimum for financial reasons – cloud computing and virtualization brought great flexibility to game server hosting but they come at a cost that needs to be rationalised. With the recent advent of PvPvE games (player versus player versus environment), it’s now more common to run an AI simulation on top of complex and resource-heavy player interactions. This chapter will go over the different ways to make the AI code performant in that context. It will describe how to efficiently limit the processing to relevant “simulation bubbles” around players, and go over various ways of applying level-of-detail techniques to the performance-intensive tasks that the AI has to run.