80 likes | 102 Views
A Test Bed for Developing Intelligent Synthetic Characters. A dozen researchers From Univ. of Michigan. Contributions. Every existing computer game is an existence proof that you can create a game without human-level AI. Demonstrate that human-level AI can make a difference.
E N D
A Test Bed for Developing Intelligent Synthetic Characters A dozen researchers From Univ. of Michigan CPSC 538a
Contributions • Every existing computer game is an existence proof that you can create a game without human-level AI. • Demonstrate that human-level AI can make a difference. • Describe a test bed for pursuing research in developing human-level AI characters within computer games. CPSC 538a
Requirement • Flexible and low-cost development --use existing game software and tools • Debugging/development environment --use existing debugging and development tools • High performance --Shouldn’t sacrifice performance in engine or interface --Support a large number of AI without sacrificing graphics performance --Able to run the game and engine on a single laptop CPSC 538a
SGIO • Soar General Input/Output • Domain independent interface between Soar and an external environment • Supports three low-level interfaces CPSC 538a
Model 1 • Drawbacks: • Overhead in SGIO to send information over the socket • Advantages: • The game can run in full-screen mode and the other machine can run full debugging environment CPSC 538a
Model 2 • Drawbacks: • can run full debugging environment but not full-screen game • Advantages: • Better performance than Model 1 CPSC 538a
Model 3 • Drawbacks: • Soar can’t display any information during runtime • Advantages: • No overhead of the interface • No overhead of the debugging environment CPSC 538a
Future Work • Create additional modules with well-defined interfaces • Create a generalized version of SGIO for non-Soar architectures. CPSC 538a