University of Science (HCMUS) - Introduction to Artificial Intelligence - CSC14003
The purpose of this project is to design and implement a logical agent that navigates through the Wumpus World, a partially-observable environment. The Wumpus World presents the following key features:
- The environment is an underground cave with a network of interconnected two-dimensional rooms.
- A room may contain a deadly pit, signaled by a perceivable breeze, or a fatal Wumpus monster, detectable via a discernible stench.
- The agent will die immediately when entering a room containing one of those harmful factors. No withdrawal is possible.
- The percepts are available in the four-neighborhood of the room containing one of those harmful factors.
- The agent has arrows to shoot in the direction he is facing.
- There is one chest of gold, located somewhere in the cave.
- Movement options: forward, turn left, or right by 90 degrees.
The primary objective encompasses locating the gold. During the journal, the agent may need to kill Wumpus to pursuit success in this environment. Figure 1 demonstrates an example of the Wumpus world with key entities.
Our Wumpus World for actual implementation slightly differs from the original problem description. The environment’s specification is as follows.
- Grid Layout
- The grid size has increased from 4x4 to 10x10.
- Elements
- Agent: The agent moves around the grid trying to achieve its goals.
- Wumpus: A monster that kills the agent if they end up in the same cell.
- Gold: The agent’s goal is to find and retrieve the gold.
- Pits: Dangerous cells that the agent falls into and dies.
- Poisonous Gas: Reduces the agent's health by 25% if entered.
- Healing Potions: Restores the agent's health by 25% when using.
- Percepts
- Breeze: Indicates an adjacent cell contains a pit.
- Stench: Indicates an adjacent cell contains the Wumpus.
- Scream: Heard if the Wumpus is killed.
- Whiff: Indicates an adjacent cell contains poisonous gas.
- Glow: Indicates an adjacent cell contains a healing potion.
- Agent's Actions
- Move Forward: Moves to the next cell in the direction the agent is facing.
- Turn Left/Right: Changes the agent's direction.
- Grab: Picks up gold or healing potions if present in the cell.
- Shoot: Fires an arrow in the direction the agent is facing to kill the Wumpus. Arrows can only hit targets in adjacent cells.
- Climb: Exits the cave (used when the agent is in the starting position).
- Heal: Uses a healing potion to restore health.
- Goals
- The primary goal remains to find the gold and return to the starting position without dying with the highest score the agent can achieve.
- Other requirements
- Recommended libraries for solving logic problems on Python: PySAT, SciPy.
- There may be any number of pits and chests of gold in the world.
- There is at least one Wumpus.
- The agent carries an infinite number of arrows.
- When the agent grabs the healing potion, the glow of that healing potion will disappear.
- The poison gas will last forever. The agent can be poisoned multiple times in the same cell.
- When the Wumpus dies. It will scream and you will know this information.
- After Wumpus dies, the stench of that Wumpus will disappear (you should update the state of map after Wumpus dies)
- The game will end when one of the following two conditions occurs:
- The agent dies (by wumpus, pit or 0% of health).
- The agent climbs out of the cave (exit the cave).
- The game will end when one of the following two conditions occurs:
The following activities can give the agent certain amounts of points.
| Activity | Points |
|---|---|
| Pick up each chest of gold | +5000 |
| Shoot an arrow | − 100 |
| Killed by the Wumpus or Fall into a pit | − 10,000 |
| Climb of the cave | +10 |
| All Agent’s Action | − 10 |
Your group must play two roles:
- Program: set up a program to build the map
- Agent: explore the Wumpus world and get the highest score possible for that world, using either Propositional Logic or First-Order Logic (or both).
Your implementation should output information about the search, including the percepts at every room the agent enters, the updates in the knowledge base after each new percept, and the action decided upon by the agent. The program should also output the score of the agent. Having your implementation generate worlds randomly can help you test your agent.
You need to create 2 classes Program and Agent in separate files.
For class Program: This class is responsible for building the map, and reporting information about the elements in the cell.
Input: the given map is represented by matrix, which is stored in the input file, for example, map1.txt. The input file format is described as follows:
- The first line contains an integer N, which is the size of map.
- N next lines with each line represents a string. If room empty, it is marked by hyphen character (-). If room has some things or signal such as Wumpus(W), Pit( P ), Breeze(B), Stench(S), Agent(A) or Gold(G). Between two adjacent rooms is separated by a dot (.)
- Input only includes Wumpus(W), Pit(P), Agent(A) and Gold(G), Poisonous Gas(P_G),Healing Potions(H_P). You need to update information about Stench(S), Breeze(B), Whiff(W) and Glow(G_L) on the map based on input data.
- For example:
-.-.W.-.P.-.-.P_G.-.- -.-.-.-.-.-.-.-.-.-
Output: The map with all infomation
Class Agent
The agent can only know the components of the cell where it is standing. The agent must go step by step and call the program to get information about the cell it is standing on from class program. From there, it makes percepts to find the direction of movement and infer objects based on the available information.
NOTE: The agent does not know the information on the whole map.
Result: a result with path for agent, game point will be displayed on screen and written in output text file such as result1.txt.
You can write output in the following format:
| No. | Specifications | Scores |
|---|---|---|
| 1 | Finish problem success | 50 % |
| 2 | Graphical demonstration of each step of the running process. You can demo in console screen or use any other graphical library. | 10 % |
| 3 | Generate at least 5 maps with difference structures such as position and number of Pit, Gold and Wumpus. | 20 % |
| 4 | Report your algorithm, experiment with some reflection or comments. | 20 % |
| Total | 100 % |
This assignment will be completed in groups, as registered at the beginning of the course. To prepare, you will need to create a folder that contains various subfolders, including source, input, output, and document. The report must give the following information:
- Your detailed information (Student Id, Full Name)
- Assignment Plan
- Environment to compile and run your program.
- Estimating the degree of completion level for each requirement.
- References (if any)
Any plagiarism, tricks, or any lie will have 0 points for the course grade.
| Student ID | Full Name | General Tasks | Detailed Tasks | Completion |
|---|---|---|---|---|
| 22127174 | Ngô Văn Khải | GUI | Implementing the GUI (Coding) | 100 % |
| Describe in detail the code of GUI (Report) | 100 % | |||
| Test cases | Describe in detail the test cases (Report) | 100 % | ||
| 22127322 | Lê Phước Phát | Algorithms | Describing in detail the algorithms of the problems successfully (Report). | 100 % |
| Test cases | Performance experiments with some reflection and comments (Report) | 100 % | ||
| 22127388 | Tô Quốc Thanh | Algorithms | Implementing the algorithms of the problems (Coding). | 100 % |
| Test cases | Performance experiments with some reflection and comments (Report) | 100 % | ||
| 22127441 | Thái Huyễn Tùng | GUI | Video demonstrating | 100 % |
| Game Play (Code + Report) | 100 % | |||
| Test cases | Generating test cases with some different attributes (position and number of Pit, Gold, and Wumpus) (Report) | 100 % | ||
| Describe in detail the test cases (Report). | 100 % |
| No. | Details | Completion Rate |
|---|---|---|
| 1 | Finish problem successfully | 100 % |
| 2 | GUI | 100 % |
| 3 | Generate at least 5 test cases | 100 % |
| 4 | Report algorithm, experiment with some reflection or comments | 100 % |
| 5 | Video demonstrating | 100 % |
- Python version: 3.10+ with Pygame graphical module.
- Operating System: the command may vary across platforms, so you need to run this program on Windows.
- Firstly, you need to set up the virtual environment in Python by using this command line below:
pip install virtualenv
python -m venv venv.
- Secondly, you need to activate the virtual environment in Python by using this command line below:
.venv\Script\activate
- Thirdly, we will install all libraries in the file “requirements.txt” using this command line below. Note that we need “cd” into the folder “Source”.
pip install -r requirements.txt
- Finally, you need to run the application by executing the command “python main.py” in the console or terminal.