The Future of Game Physics

By: Andy Thomason,
SN Systems, Sony Computer Entertainment

This discussion builds on the speaker’s experience of game physics from the 1970’s to the present and investigates current research to try to establish which way interactive entertainment may go in the future. There is a brief discussion of the state of the art in physics simulation, such as sequential impulse solvers and hash-based broadphase collision detection. We investigate biomechanics simulation, the use of AI to control simulated musculature and game environments derived from the real world as potential targets for future development.

Bio: Andy Thomason works for SN Systems, now a part of Sony Computer Entertainment, developing the compilers and tools for the PlayStation(R) range of consoles. Andy started writing games, operating systems and compilers in the 1970’s and spent all his pocket money on Z80s, 2708s and Veroboard. In the 80's he wrote the 3B2 publishing system, before returning to games and technology to work with Sinclair Research, Satchi and Satchi, ARC and Samsung on parallel computing operating systems and hardware. In the 90's he was a member of the Psygnosis technology group, moving to Rage Games for "B17 The Mighty Eighth", Confounding Factor for "Galleon". Recently he has been writing compilers and optimising games and middleware, such as Killzone 2, Fallout 3 and Unreal. When not fixing compilers, Andy is a keen mathematician and dad.

Learning to Play: Machine Learning and Computer Games

by: Thore Graepel
Microsoft Research

Abstract:  In the Applied Games group at Microsoft Research Cambridge we are working on the application of machine learning to computer games. I will talk about the motivation for this work and describe three projects in more detail. In the Drivatars project our team built the driving AI for the commercial racing game Forza Motorsport for Xbox. What sets this AI system apart is its ability to learn the driving style of each individual gamer through the application of machine learning methods and provide a “cloned” AI driver that mimics the player’s style in game. In the Tao Feng project, we applied the paradigm of reinforcement learning to train an AI system to be able to compete in a fighting game. The system learns to fight by playing against the built-in AI and by adapting its policy from observed state-action-reward sequences. Finally, in the TrueSkill project, we addressed the problem of estimating the skills of gamers in particular games or game modes to be able to match them into balanced, fair, and fun online matches. TrueSkill is based on a principled Bayesian modelling approach and can be viewed as a generalization to the Chess rating system ELO to the multi-player/multi-team scenario. For all three projects, I will present motivation, conceptual and mathematical background as well as illustrations/demos of the systems at work.

Bio: Thore Graepel studied physics in Hamburg, London and Berlin. He received both his diploma in physics and his PhD in statistics from the Technical University of Berlin. He worked on machine learning and large scale optimisation at the Swiss Federal Institute of Technology Zurich and on kernel methods and learning theory at Royal Holloway, University of London. Since 2002 he has been a researcher in the Machine Learning and Perception group at Microsoft Research Cambridge where he co-founded the Applied Games group in 2006. The group’s mission is to apply machine learning to games. These include both recreational games such as Go, poker, and video games as well as abstract decision games such as auctions and negotiations played in the real world. The group developed the TrueSkill™ system used millions of times each day in Microsoft’s Xbox Live service for ranking and matchmaking players according to their skills and which is also the backbone of the ranking system of the blockbuster game Halo 3. APG also developed the Drivatar™ AI for the Xbox title Forza Motorsport™. Thore has published over 50 technical papers on topics around machine learning including probabilistic models, statistical learning theory, kernel methods, clustering, stochastic optimisation and reinforcement learning with applications including computer go, video games, computer networks and ranking. He is a member of the editorial board of Springer’s Machine Learning Journal and of the IEEE Transactions on Computational Intelligence and AI in Games.

KEEPING YOUR GAME

Presented by: Thomas Coughlin, PhD
IEEE CE society Distinguished lecturer on Storage

Abstract:  Immersive and realistic gaming experiences require fast access to rich content.  This places significant performance and capacity requirements on storage devices and systems.  Local storage used in game systems may create a new type of storage hierarchy matching game requirements with costs and performance to create the optimal local game experience. 

Likewise on-line games are creating new opportunities for storage as a service (SaaS) geared to the needs of the connected gamer.  Stereoscopic technology is about to explode into the home display market and implementation of stereoscopic games as well as even higher resolution content (such as Ultra-HD and virtual reality) will create new requirements for storage devices.  Social networking and user generated content (UCG) inserted into games could soon change the very nature of games and make very compelling and personalized game experiences.  Use of UGC in games may require implementation of advanced automated metadata generation and indexing as well as standards enabling recognition and character modeling and insertion based upon this personalized content.  

This presentation will review how digital storage is used in modern games and game systems. It will explore a possible evolution of game storage devices and how new developments in digital storage and storage implementation in game systems combined with a greater variety of user and commercial content could create whole new worlds of game experiences.

Biography:
Tom Coughlin, President, Coughlin Associates is a widely respected storage analyst and consultant.  He has over 30 years in the data storage industry with multiple engineering and management positions at high profile companies. 

Tom has many publications and six patents to his credit.  Tom is also the author of Digital Storage in Consumer Electronics:  The Essential Guide, which was published by Newnes Press in March 2008. Coughlin Associates provides market and technology analysis (including reports on several digital storage technologies and applications and a newsletter) as well as Data Storage Technical Consulting services. 

Tom is active with SMPTE, IDEMA, SNIA, the IEEE Magnetics Society, IEEE CE Society, and other professional organizations.  He is a senior member of the IEEE and a member of the CE Society Adcom as well as Central Area Chairman of Region 6.  Dr. Coughlin is the founder and organizer of the Annual Storage Visions Conference (www.storagevisions.com), a partner to the annual Consumer Electronics Show as well as the Creative Storage Conference held before the NAB show (April 19, 2009, for more information go to www.creativestorage.org). Tom is also the chairman of the annual Flash Memory Summit.  He is also a Leader in the Gerson Lehrman Group Councils of Advisors and a member of the Consultants Network of Silicon Valley (CNSV).  For more information go to www.tomcoughlin.com.

Artificial Intelligence
and Games Research
Network