#bwapi

[hackthechad]

Case-Based Reasoning AI Research Updates

If anyone is wondering what ever happened to my CBR AI Research, it is being presented tomorrow at the ICCBR 2010 conference in Alessandria, Italy. If you hop on a plane right now you might get there in time to see it! Here is the CBR Workshop for Computer Games schedule: ICCBR Conference 2010 Website

TimeTopic09:15 - 09:30 Introduction and Goals of the workshop09:30 - 10:10 Invited Talk: Artificial Intelligence for Games: Challenges and Opportunities Pedro A. Gonzalez Calero 10:10 - 10:30 Using Automated Replay Annotation for Case-Based Planning in Games Ben Weber and Santiago Ontañón10:30 - 10:45 Coffee Break10:45 - 11:05 Toward a Domain-independent Case-based Reasoning Approach for Imitation: Three Case Studies in Gaming Michael Floyd and Babak Esfandiari11:05 - 11:25 Feature Selection for CBR in Imitation of RoboCup Agents: A Comparative Study Edgar Raul Acosta Villaseñor, Babak Esfandiari and Michael W. Floyd11:25 - 11:45 Meta-Level Behavior Adaptation in Real-Time Strategy Games Manish mehta, Santi Ontanon and Ashwin Ram11:45 - 12:05 Online Micro-Level Decision Making in Real-Time Strategy Games: A Case-Based Reasoning and Reinforcement Learning Approach Chad Mowery, Nathan Spencer and Isabelle Bichindaritz12:05 - 12:25 MMPM: a Generic Platform for Case-Based Planning Research Pedro-Pablo Gómez Martín, David LLansó, Marco Antonio Gómez Martín, Santiago Ontañón and Ashwin Ram12:25 - 12:30 Closing Remarks

Also, if you are a Twitter user you can follow the conference progress using: #iccbr. OR you can be lazy and click this link to go there from here. I will post more about my research on this topic after it has been printed.

[hackthechad]

Case-Based Reasoning and Reinforcement Learning in Real-Time Strategy Games.

I just gave a presentation on using Case-Based Reasoning and Reinforcement Learning in RTS games to facilitate micro-level decisions back on the 12th at the University of Washington Tacoma campus, so I thought I would share some of the AI concepts used in my work. Game: Starcraft with the Brood War expansion (Blizzard TM). Tech: BWAPI Which is a lovely library that provides hooks into Starcraft and allows you to provide a custom AI dll to inject into the game at runtime using just about any programming language you prefer (I used C++). Find it here: http://code.google.com/p/bwapi/ So what is Case-Based Reasoning (CBR)? CBR is a significantly different approach to solving problems when compared to more traditional AI techniques such as decision trees, state-based machines, and various expert systems. That is, CBR records and uses specific knowledge from previously experienced problem situations. Thus, problems are solved by recalling these previous experiences (cases) and applying them in new the new problem context. This kind of problem solving is very interesting as there are many strong similarities with our own natural cognitive abilities. We (humans) often learn by example and experience. The recall, adaptation, and application of these experiences is almost identical to the fundamental processes in CBR. And what about Reinforcement Learning (RL)? RL is used as a means for our AI to adapt it's experiences (cases) and make certain generalizations about them in order to extend their usage across larger problem situation domains. This way, our AI must gain experience by playing games of Starcraft so that it can learn how to solve problem situations during game-play. Additionally, it also has to learn  how to play a game of Starcraft (i.e. what it means to win and how it must go about doing this). RL simply uses reward and punishment functions following a feedback loop during a game in order to provide a positive or negative reinforcement to a recalled case in hopes to further refine the solution contained within it. I don't claim to be an expert on either of these topics as I am still learning about them through this project. If you want more in-depth information here are a few links: http://www.aaai.org/AITopics/pmwiki/pmwiki.php/AITopics/CaseBasedReasoning http://www.aaai.org/AITopics/pmwiki/pmwiki.php/AITopics/MachineLearning