This is an ongiong project, which aims to provide a versatile platform for testing reinforcement learning algorithms on a quadrotor system. There are in-build PID controllers that can be used for attitude, altitude and position control of the drone. The simulation platform is built on matplotlib.
You will require the keyboard library for pausing, unpausing or reseting the simulation from the keyboard.
pip install keyboard
Make sure that you have both the files in the same folder. After that, simply run
sudo python main.py
This will open up a matplotlib animation window, where you can see the drone stabilizing at a given point.
The mathematical model of the drone is obtained from Wilselby's Masters thesis. Here are the links to his personal blog, and his github website:
https://www.wilselby.com/research/arducopter/simulation-environment/
https://github.com/wilselby
The differential equations are solved using Euler's method. However, if you change the setpoints arbitrarily, the simulation may go crazy. This might be due to the Euler's method of integration. Other methods will be tested, and updated in the file.
Any changes made will be updated to this repository.