[Add] GA implementation, pypi project structure, trainer gui

.gitignore

@@ -3,5 +3,5 @@ __temp__
 __temp__.tga
 .idea
 .venv
-.pickle
-.png
+.png
+/robo-env/

README.md

@@ -1,72 +1,59 @@
-
 ### Robot Co-optimization using planar linkage
 The base code implements the planar linkage kinematics in `linkage.py`, and the linkage can be installed on a robot and simulate to test the robot performance using `linkage_physics.py`.  Finally, we implemented a basic linkage optimizer using simulated annealing in `optimizer_anneal.py`.
 
-#### Install
-
-The code has been tested on Python 3.6 and 3.12. 
-
-To run the code, you first need to install Anaconda, which can be downloaded from https://docs.anaconda.com/anaconda/install/
-
-Once Anaconda is installed, you can open the anaconda and type the following commands in order:
-
-1. now, you can create your conda environment, named "box2d-py"
-
-`conda create -n box2d-py`
-
-2. remember every time you activate the environment, before you use it.
-
-`conda activate box2d-py`
-
-3. once you activate the environment for the first time, you need to install the following packages.
-
-`conda install conda-forge::box2d-py` (box2d-py (https://github.com/pybox2d/pybox2d). Fortunately, you can install it via Anaconda.)
-
-`pip install pygame`
-
-`pip install numpy`
-
-`pip install vapory`
-
-4. run the Python scripts one by one
-
-`python linkage.py`
-
-`python optimizer_anneal.py`
-
-`python linkage_physics.py`
-
-### About linkage_physics.py
-
-To run linkage_physics.py, you need the 'best.pickle' file that contains the design from optimizer_anneal.py. That said, before you run optimizer_anneal.py, you do not have it yet. 
-
-If you want to simply test the linkage_physics.py, instead, you can replace 
-
-`robot=create_robot('best.pickle', sep=5.)`
-
-with 
-
-`link=Linkage.createSimple()`
-
-`robot=create_robot(link, sep=5.) `
-
-Once you run linkage_physics.py, you can control the robots via
-
-press 'SPACE' to turn on/off simulation
-
-press 'RCTRL' to turn on/off robot motor
-
-
-### Installing Povray on Windows
-
-On windows, the default povray does not work. We need a special version of povray that is independently compiled called MegaPov.
-I have included a version of this MegaPov in the project, i.e., the megapov.rar. To enable povray on windows, unzip this folder to any directory.
-Let's say you have a povray.exe located in "C:\Users\Admin\AppData\Roaming\POV-Ray\bin".
-After this step, you need to add this to your system path as follows:
-![alt text](https://github.com/dyingbrain/RoboCoOpt/blob/main/sys.png)
-
-#### Note that you might need to restart the system to let the setting enabled.
-
+### My Contribution
+
+#### Day 0
+- Cloned the repo, created a venv for the project, installed the required packages, solved various dependency issues.
+- Since I'm not using conda, my resolution to the issues associated with `Box2D` was to install some additional dependencies like `swig` and `box2d-py` using `pip`.
+- Ran `linkage.py`, `optimizer_anneal.py` and `linkage_physics.py` to verify the installation and the working of the code.
+- Studied the codebase to understand the implementation of the planar linkage kinematics, the simulated annealing optimizer, and the linkage physics simulation.
+
+#### Day 1
+- To see the best possible results being generated by the optimizer for 1000 steps, I  automated the process of running the optimizer and kept a track of the best results by updating the `.pickle` file if the current score is greater than the high score.
+- The best fitness score I got after 8 iterations was `~10.77` after 28 iterations of the optimizer.
+- I created a simple GUI around this functionality to keep track of all the results and the best result using `PyQt5`.
+
+<p align="center">
+  <img src="https://github.com/user-attachments/assets/6f84339c-da97-4221-b358-d3ee2c68577a" alt="record">
+</p>
+
+https://github.com/user-attachments/assets/9985374c-505e-48f7-9b4a-3bfdfeb526aa
+
+#### Day 2
+- Let the tool run for the entire day, achieved a fitness score of `~25.5` using Simulated Anenaling over multiple iterations using the trainer.
+- Implemented genetic algorithms by Tina Wu.
+
+<p align="center">
+  <img src="https://github.com/user-attachments/assets/110c3e45-b22a-4aef-ac7d-98d572f29fd5" alt="record">
+</p>
+
+#### Day 3
+- Fixed several bugs associated with genetic algo, improved upon the code design.
+- Integrated `.pickle` export functionality to visualize the results in `pygame`.
+- Current fitness score is in the ballpark of `2-5`
+- Need to work on defining our own `crossover` and `mutation` implementation for genetic algorithms to function correctly.
+- Dk why the wheels are suddenly so big with no change to the codebase 💀
+
+<p align="center">
+  <img src="https://github.com/user-attachments/assets/ebd2197b-2055-4b4a-b5cd-c93a0d1dc76e" alt="record">
+</p>
+
+#### Day 4
+- Implemented parallel processing for the genetic algorithm to speed up the process.
+- Fixed the issue with the wheels being too big.
+- Implemented the custom `crossover` and `mutation` functions to improve the results by the genetic algorithm.
+- Fitness score of `~24.05` achieved after over 100 generations with a population size of 100 in each generation.
+
+<p align="center">
+  <img src="https://github.com/user-attachments/assets/30ede664-16e4-45f1-bad2-3a47cdbfc204" alt="record">
+</p>
+
+#### Day 5
+- Converted the project structure into that of python package.
+- Extended the trainer gui to include the genetic algorithm functionality as well as accepting parameters for both algorithms.
+
+<p align="center">
+<img width="411" alt="image" src="https://github.com/user-attachments/assets/b2cc8722-350b-4bc7-849a-2a3558e9778b">
+</p>
 
-Now you can press "LSHIFT" in the window to have the povray render a nice image for you, an example is as follows:
-![alt text](https://github.com/dyingbrain/RoboCoOpt/blob/main/frm.png)