Setup

• Run python setup.py install. This file can be found in gym-maze.
• Install the requirements from gym-maze/gym_maze/envs/requirements.txt. Use pip install -r requirements.txt

Generating a Maze

• Run maze_generator.py which can be found in gym_maze/env.
• rescue_item_locations is overriden internally and the rescue items will be randomly placed.
• The default maze size for evaluation will be 10x10, but feel free to experiment with other sizes by adjusting maze_size.
• Below is a snippet of the main function.
• The resulting maze will be a .npy file named sample_maze.npy, saved in gym-maze.

if __name__ == "__main__":
    # DO NOT CHANGE
    # USE ONLY FOR GENERATING RANDOM/SAMPLE MAZES
    while(True):
        maze = Maze(maze_size=(10, 10), rescue_item_locations=[(10,10)])
        is_validated = validate_maze(maze.maze_cells)
        if is_validated:
            break
    print(maze.maze_cells)
    np.save("../../sample_maze.npy", maze.maze_cells)

Running Locally

• Run python local_solver.py.
• Write any agent_id. This is used only when submitting.
• Make sure you add the path of the maze you want to try in the main function.
• Riddle solvers will be imported from riddle_solvers.py, so make sure to implement your riddle solutions there.

Submission

• Make sure that you are using the correct IP and endpoints.
• Make sure that you are using the correct agent_id.
• Riddle solvers will be imported from riddle_solvers.py, so make sure to implement your riddle solutions there.
• Run submission_solver.py

Endpoints used in submission

Start by calling /init endpoint to initialize a game, then use /move to move inside the maze. To solve a riddle call /solve. After finishing the attempt call /leave

/init

POST

Description:

Initialize an agent, this endpoint should be used for first time connecting to the server

Request

{ "agentId" : "ABC123" }

Responses

/move

POST

Description:

Send a move action

Request

{ "agentId" : "ABC123", "action" : "N" }

Responses

/solve

POST

Description:

Solving a riddle

Request

{ "agentId" : "ABC123", "riddleType" : "captcha", "solution" : "solution" }

Responses