Control Experiments

This repository outlines various control experiments using modern control theory and state space controller design.

The results are simulated using pymunk (a 2D physics simulator) and pyglet

Requirements

Modern version of Python (developed on Python 3.7.1).

It is highly recommended that you create a Python virtual environment before installing the python libraries.

To run the simulations in a game engine:

pip install pymunk pyglet

To run the notebooks:

pip install numpy scipy matplotlib jupyter sympy

Running the Samples

Simulations

Inverted Pendulum on a Moveable Cart

Controlling an inverted pendulum using a moveable cart. The simulation commands the cart to various positions every few seconds while still keeping the pendulum stable in a vertical position.

python invpend.py

Double Inverted Pendulum on a Moveable Cart

Controlling a double inverted pendulum using a moveable cart. The simulation commands the cart to various positions every few seconds while still keeping both pendulums stable in a vertical position.

python dinvpend.py

Quadcopter in 2D space

Controlling a quadcopter in 2D space. The simulation commands the quadcopter to several waypoints every few seconds.

The trajectory is calculated using different trajectory generators: minimum snap, jerk, acceleration and velocity. After each trajectory finishes, a couple seconds will pass before the quadcopter moves onto the next trajectory generator. All the trajectories are plotted ahead of time using various shades of grey. The actual trajectory is rendered in yellow. Trajectory points are rendered in red as small crosses.

Change the points array for the desired waypoints, and update the average velocity to automatically calculate the desired waypoint arrival times.

python quadcopter2d.py

Quadcopter in 3D space

At this time, there is only a notebook deriving the equations of motion and some simple simulations using PD control. In future I hope to add a 3D simulation that uses a minimum snap trajectory generator and PID control.

Notebooks

Start Jupyter:
jupyter notebook and go to http://localhost:8888 and load the notebook you desire

The notebooks derive the motions of equation using sympy (symbolic math library) and then design controllers. The results from the notebooks are used to test the results in pymunk and pyglet.