Warning
This page is located in archive. Go to the latest version of this course pages.

Search (1st assignment)

You will program the A* algorithm in order to search through the given state space.

Task

Program an A* algorithm that finds the optimal path to the desired end state. If there is no optimal path, the returned value should be None. You should use the maze environment kuimaze.InfEasyMaze. For fulfilling the task, you need to select and implement a heuristic function. The impact of selecting different functions is well explained in link.

Learning outcomes

This task will teach you about:

  • Time complexity
  • Space complexity
  • Resulting path quality

How to

  1. Create the file agent.py according to the guidelines.
  2. Use the testing environment supplied in the ZIP kuimaze_search.zip file for testing before submission.
    1. Extract the ZIP file in the folder with your project.
    2. You need to have the following python3 packages: AI-gym, numpy and pillow, for proper use of the scripts. You can get these packages by running these commands from your terminal (preferably from a virtual environment, or a conda environment):

# On Linux
pip3 install --user gym numpy
sudo apt-get install python3-pil.imagetk
 
# On Mac with Homebrew package manager
pip3 install gym numpy
pip3 install Pillow
brew install python-tk
 
# On Windows
pip3 install gym numpy
pip3 install Pillow

  • The code must be compatible with Python 3. Otherwise it is possible that the automatic evaluation system will malfunction!

All the methods and functions in kuimaze_search.zip are documented in the subfolder /kuimaze_doc/ or available at this link.

Agent class and its methods

The ZIP file contains the file easy_example.py. This file serves as a tutorial to the environment as seen on this code snippet with the basic functions:

import kuimaze # package import
MAP = 'maps/normal/normal9.bmp'
env = kuimaze.InfEasyMaze(map_image=MAP) # create the environment
observation = self.environment.reset() # returns start_pos, goal_pos
positions_with_costs = env.expand(position) # returns [(pos, cost)] list

Implement your agent in the file agent.py as the class Agent, inheriting from the class kuimaze.BaseAgent.

The final agent must have the following methods:

Method input parameters output parameters note
__init__ environmnent none Agent initialization.
find_path none path Generates the path. The method returns a list of coordinates for the path. It must start with the starting position and end at the goal position [(x1, y1), (x2, y2), … ]. If there is no path, it should return None.

Evaluation

The evaluation splits into two:

  1. Minimal solution that returns a correct path. If there is no path, it returns None.
  2. Manual evaluation of the code and its form (clean code).
Evaluation min max note
Algorithm quality 0 5 Evaluation given by the automatic evaluation system.
Code quality 0 2 Comments, code structure, cleanliness, proper naming…

Algorithm quality:

  • tested on multiple mazes with various size and complexity
  • is the path valid? - does not cross walls and is continuous on the 8-neighbourhood
  • is the path optimal?

Code quality:

  • suitable comments or the code is so clear that it does not need comments
  • reasonable length of methods
  • variables and functions are properly named and this naming helps understanding and reading the code
  • no repetitive code (copy-paste in the same code)
  • saving processing time and memory (i.e. no unnecessary nested loops)

You can follow PEP8, although we do not check all PEP8 demands. Most of the IDEs (certainly PyCharm) point out mishaps with regards to PEP8. You can also read some other sources for inspiration about clean code (e.g., here) or about idiomatic python (e.g., medium, python.net).

Submission

The date for submission is visible in the Upload system.

* Upload a ZIP archive with the module agent.py and any other possible modules you created to the Upload system. All the files must to be in the archive's root folder! The archive cannot contain any other folders!

Screenshots

Possible final screenshot:

courses/be5b33kui/labs/search/start.txt · Last modified: 2023/02/27 10:45 by gamafili