Home

QuakeAI

Quake A.I. is an on-going research project with the goal of creating challenging games customized for every player. It will be applied AI techniques for analyzing players skills and for deciding how the AI should behave accordingly. For this purpose we have developed a Quake3 mod which consists on a weapon-based match between two players in a map with a great variety of collectable items that they need to take in order to increase their chance of winning. We have chosen this particular shooter game as it provides a fitting environment for designing and implementing AI algorithms. This new AI approach can be broken down in the following main development phases:

• Modeling phase. We define the data structures necessary to represent a discrete approximation of the virtual world and implement algorithms that create the discrete world via physics simulation.
• Decision-making phase. We implement a runtime decision-making system for AI duels that simultaneously simulates both NPC player actions using heuristics to represent numerically how good or bad these actions are. Each action is executed in parallel through a simulation and once all the simulations have been performed, the decision-making algorithm chooses the most optimal solution for each player. The decision will be updated and synchronized with the implemented AI agent that carries out the decision-making plans.
• Challenging phase. We propose a post-game processing algorithm that analyzes the recorded game actions of the human player using the same AI model and decision-making process. The offline process will numerically evaluate each player’s decision by comparing them with other alternatives. Other player skills such as environment awareness, aiming or reaction time will be also considered and numerically evaluated.

We have written documentation in the wiki section that is meant for readers with a technical background in computer science, especially in the artificial intelligence field.