Comprehensive Survey of peer-to-peer architecutres used for Massively Multiplyaer Games

We did the most comprehensive study of different peer-to-peer architectures proposed for supporting massively multiplayer online games. We provide a comparison of different architectures as well as the necessary background qnd a range of design principles (interest management, structure, replication, update dissemination & multicast, fault tolerance, persistence, cheating, industry adoption, etc). Submitted to ACM Computing Surveys. 44 pages and 2 years later: ACCEPTED

Prediction of player's behavior and applications

By modeling player's interests we are able to predict their movements and behavior in order to improve the accuracy of algorithms such as dead reckoning. Players' movements and interactions are not only a function of their current speed and game physics but are mostly influenced by players' interests. A wounded player is likely to pick up a weapon and a player being attacked will flee the attacker. We model these interests in our system and then use this information in order to better predict how players interact with the game world. These predictions are then used to greatly reduce network requirements and provide a better gameplay experience. See our paper at Multimedia System Journal (MMSJ) and NetGames'11

Cheat resitatnt scalable distributed multiplayer games

Using randomized indirect subscription management and mutual verification we provide a distributed protocol that can be used to safely scale multiplayer games. In this protocol we divide players into differnet sets with different requirements therefore allowing us to provide scalability for the game. Randomized proxies and mutual verifications are used to ensure that players do not gain an unfair advantage and to hide unnecessary information from the players. Check out our videos showing our implementation of interest sets. See our paper at NetGames'11 and the technical report. ICDCS paper submitted..

Gossip-based neighbor discovery

We are working on gossip-based and area-cast neighbor discovery for games. These techniques rely on unstructured P2P networks for a discovering a continous game world. Currently on hold.

Realistic NPC players using interest modeling

By using interest modeling and trace analysis (see first project) we suggest new methods that can be used by the AI to exhibit more human like behavior.

(credit: zazzle)

Load balancing in multi-server game environments

We are starting a project on analysis of different load balancing techniques for distributed game server environments. Just started...

Mammoth pSense integration

On hold.

Check back for more projects (trace generation, mobility, etc)...