Examples

These examples demonstrate how to model different search problems with AI Gym.

Vacuum world

VacuumWorldGeneric.py receives a text file describing dirty rooms and the robot's initial position. In the map, 0 means clean and 1 means dirty.

0;1;1;1
0;0;0;0
1;1;1;1

Run the example with the configuration file, row, and column:

python docs/src/VacuumWorldGeneric.py configuration.txt 0 0

The available actions move left, right, up, or down, or clean the current room. A valid solution reaches a state in which every room is clean.

Another configuration can start the robot on row 2, column 3:

0;1;1;1
0;0;1;1
1;1;1;1

python docs/src/VacuumWorldGeneric.py configuration.txt 2 3

View implementation

U2 bridge problem

The four members of U2 must cross a bridge in 17 minutes. It is night, they have one flashlight, and no more than two people can cross together. Bono, Edge, Adam, and Larry take 1, 2, 5, and 10 minutes respectively. A pair moves at the slower person's speed.

The objective is to move everyone and the flashlight to the other side with the lowest total cost.

View implementation

8 Puzzle

The 8 Puzzle example uses A* to arrange numbered tiles into the target state.

View implementation

Noughts & Crosses

The Noughts & Crosses example models a complete 3×3 tic-tac-toe game for MinMax. X is represented by 1 and maximizes utility; O is represented by -1 and minimizes it; empty squares use 0.

Each state stores an immutable board and the next player. successors() creates one state for every empty square. cost() returns a large terminal utility for a win and a line-based estimate at the depth limit.

Run the example from the project root:

python -m docs.src.NoughtsNCrosses

To display the decision tree, change the final call to play(trace=True). MinMax displays two levels by default while preserving the complete selected path. Increase trace_max_depth only for small game trees.

View implementation