This thesis may be interesting for developers within video game industry, people who design the behaviour of agent systems, or those interested in complex robotic control systems. What potential fields give to the world of AI is a new way to implement the behaviour of the agents, giving weights to all the elements in the world, so the agent can decide if making some action deserves the cost, the combination of several potential fields is what makes this technique so efficient. An agent was created applying this technique. Implemented as a client application, the agent used the Quake protocol to connect and communicate with the server. Some tests were run on the model concerning the different functionalities a gamer agent should implement. The model turned out to be a success in a virtual environment like quake 2. Taking care of all the entities in the map and environmental factors, the agent is able to keep his state in balance, move to the most interesting positions and be able to face combat situations. The implementation is clean and mathematical oriented. Using algorithms as the heart of the model, we make the code lightweight in comparison with old bot models. / The purpose of this thesis was to test an artificial intelligence technique called "Multiple Potential Fields" in the programming of an agent playing in a multiplayer game as Quake 2. The results stated that the technique is really interesting to apply in an virtual real-time environment, having some clear advantages compared to other techniques. / hectorvcaz@hotmail.com

In most games today the movements of computer players is controlled by a web of waypoints that have been predefined by the creators of the map, or automatically generated by an application on beforehand. This data is precompiled to save precious CPU cycles for the player of the game, but what if we want these computer players to be able to explore how to get around the map by themselves? How could this be done and how would the result change if we had several of these players cooperating to explore the map? We explore some of the possibilities for exploration and discovery of maps using approaches that is often found in robotics. We also look at what happens when there is a penalty for communicating between these computer players, and what effect different amounts of map coverage will have on the performance of them. By setting up a test environment inside an existing commercial computer game we developed a client side bot system that made it possible for us to test different parameters and settings. Based on the analysis of the test results we propose ways to describe and predict the effects of the number of bots and the communication rate, and we describe several ways of how to further advance on these ideas and experiments.