This program simulates agent path planning using A* and Q-Learning in a 2D grid.
- Verify the .NET framework version 4.5 or higher is installed.
- Run in either "Debug" or "Release" mode in Visual Studio.
- Click the button "Load Grid File…" and select a well-formed CSV file. For the purposes of this assignment, a well-formed CSV file will have exactly 10 lines with each line containing 9 commas. Quotation marks should not be used in a file. The symbol definitions are:
| Symbol | Meaning |
|---|---|
| 0 | An obstacle in the 2D grid. |
| 1 | The agent in the 2D grid. |
| 2 | The reward in the 2D grid. |
| Here is an example of a well-formed grid map CSV file with various obstacles. The agent is in the top left corner and the reward is in the bottom right corner: |
1,,,,,,,,0,
,0,,0,,,,,,
,,,0,,0,,0,0,0
,,,0,,,,,,0
,,0,,,,,,,0
,,,,,0,,,,0
,,,0,,0,,0,,
,0,,0,,0,,0,,
,,,0,,0,,0,,
,,,0,,,,0,,2
- Select either "Q-Learning" or "A*" and click "Start". The default algorithm is A*.
- For A*, the best path is found and will be automatically displayed after the search. For Q-Learning, the Q-Table is trained. After training, click on a cell to have the agent navigate the grid from the clicked cell to the reward using the max exploration policy.