To create both grounds for experimentation and demonstration a game environment was designed and built in 3d game studio was designed to both immediately and clearly bring to life the operations and thinking of the Knowledge dll.

The results from this experiment are described in part four.

The game is a large room filled with various 'machines' that can be operated by simple commands - such that they may be operated by the Knowledge dll. These machines are various obstacles that the human player must get past. These include:

• 3 x heavy partitions
• 3 x caged monsters
• 1 x stairs in put
• 1 x bridge over pit

Map Layout

Game Screenshots

The opening view (gamma+160%), turning right takes the player in the large room	View of the main room (gamma+160%)
The bridge in motion and the stairs in rest position (gamma+160%)	A creature emerging from her cage (gamma+370%)

Gameplay

The goal of the player is to successfully reach the other side of the main room. The Knowledge dll has the freedom to operate the machines at set periods. The partitions, stairs, and bridge are all components that can be used to prevent the player from advancing across the room. The cages each contain creatures that are programmed to follow the player until they catch him or her.

When the partitions, bridge or stairs commands are executed, each is closed/moved, but only for a limited period of time.

If a player is caught, then they have lost. If a player crosses to the other side of the room, they have won.

Whenever a win or lose state occurs, the game is reset to the beginning, and the Knowledge dll is informed.

Therefore the challenge for the Knowledge program is to identify strategies that will successfully hurt the player. For example, closing a partition and letting a creature loose. The player, having nowhere to run, will be caught.

Exact Details

Partition A (id: 1001), B (id: 1002), and C (id: 1003)

Closes in ~2.525 seconds (fast to deal with catching the user before they reach the partition)
Stays closed for ~22.5 seconds
Opens in ~4.75 seconds

Bridge (id: 1008)

Closes in ~9.25 seconds
Stays closed for ~22.5 seconds
Opens in ~9.25 seconds

Stairs (id: 1004)

Moves in ~ 1.875 seconds
Stays closed for ~22.5 seconds
Moves back in ~5.625 seconds

Cage A (id: 1005), Cage B (id: 1006), and Cage C (id: 1007)

Opening time: ~5 seconds

Walking time

Partition to partition is approximately ~3 seconds (enough time to close a partition before the player reaches either the first or the second.

Connection to the Knowledge DLL

Data can be passed to the Knowledge dll at different points of the game, changing the way it learns. The quantity of information fed to it also changes the complexity and skill involved in the learning process.

There are two techniques which define different techniques the game uses for learning.

Event State Reaction

The Event State reaction allows the Knowledge dll to operate two machines every time the player enters a new region of the room. This means that there has to be some relation of where the user is to exactly what devices are controlled.

Five Second Total Plan

In this state the software is only guessing sequences of events that will lead to a win-state.

Every five seconds the Knowledge dll is permitted to make a move in the game.
No feedback is recorded except a win or lose state.