Click on images for full size.
Click on images for full size.
Skookum is a classic adventure game with a few modern twists. Skookum allows a player to explore a classic Midwestern Lumber-town of the 1890’s in full 3D. While driving into town to deliver a package of saw blades for your employer, the Western Supply and Fur Trading Company, in one of the first cars ever produced, the 1896 Duryea, you run off the road due to mechanical failure. This provides you with the starting point of your adventure and your first two objectives: deliver your package of blades to the local lumber mill foreman and then get your motor wagon fixed so you can get back to Chicago. The only thing is that something is not quite right in this small town...
Reference System: Intel Core 2 Due @ 2.1GHz, 2 GB Ram, DirectX 9, Windows Vista, NVIDIA GeForce 7900 GS, Pixel and Vertex Shader Model 3 required.
Skookum takes advantage of three main types of artificial intelligence techniques. A unique hybrid dialogue system was developed integrating traditional dialogue trees structures into an AIMLbot [Tollervey] utilizing the AI Markup Language [Wallace]. This system performs symbolic pattern matching for conversation query and response and includes keyword learning through the discourse to improve continuity. Grounding free-form text discourse with a dialogue tree keeps the player focused, but also allows for improved interactiviy. The Common Game Understanding and Learning (CGUL) Toolkit [Youngblood, Dixit, and Hale] was used to decompose world geometry into navigable regions for bots annotated with geometric traversal information. Graph connectivity learning was performed offline to produce a navigation mesh. The SARGE (SPSS Automated Region and Gateway Extractor), DEACCON (Decomposition of Environments for Automated Creation of Convex-region Navigation-meshes), and SSPS (Static Spatial Perception Service) elements of CGUL were used to produce this knowledge for the game bots. Finally, standard A* search [Hart, Nilsson, and Raphael] was used to for NPC path planning for movement. Simple finite state machines were used to control the dialogue and characters of the game.
The CGUL Toolkit can be found online at playground.uncc.edu/GameIntelligenceGroup/CGUL.