jataruku

A PID controller written in Python3


Licenses
CERN-OHL-P-2.0/SGI-B-2.0
Install
pip install jataruku==0.1.1

Documentation

jataruku

This project provides a PID controller written in Python3 in contrast to the orignal. The original is made by Jeremy Bornstein This is based on the Arduino PID code written by Brett Beauregard.

Brett's code is the result of much research and thought on his part, which he has generously documented. Most of what I did is translate it into a form which I hope can be easily used if you are implementing a system in Python that wants to use PID control.

How to Use

The module exports only one object, a class called PID. The best way to learn how to use it is to look at the test.py file (see below).

The second-best way to learn how to use it is to keep reading this section.

The PID object must be constructed with the following parameters:

  • input: a lambda which reports the input value, e.g. the temperature of the object the temperature of which is being controlled

  • output: a function which controls the output setting, e.g. whether or not the heater is turned on or not. it takes a single parameter, which is the value that the output object should receive

  • setpoint: the target value for input, e.g. the desired temperature of the object in question

  • kp, ki, and kd: the three PID parameters. (If you don't know what they are you should go read about how PIDs work, though you may be able to get away with using defaults based on the test code herein.)

  • direct: True if an increase in the output value will produce an increase in the input value, False otherwise

pid = PID(input, output, setpoint, kp, ki, kd, direct)

After the object has been created, turn PID control on with pid.auto = True, and set the range of your output control with pid.set_output_limits(min_out, max_out). At that point you can just call pid.compute() as frequently as possible, preferably via an interrupt of some sort. The frequency required for good results will depend on the characteristics of your entire system, but the default update frequency of 100 milliseconds is probably a good place to start. If the system is working, then the input value will approach the set point and stay there.

You can change the system back and forth from automatic to manual mode at any time by setting pid.auto to False (for manual control) or True (for automatic).

If you want to change the sample rate, set pid.sample_time to a number of milliseconds. The default value is 100. If you want to change the set point, just set pid.setpoint, and if you want to re-set any of the PID parameters you can call pid.set_tunings(kp, ki, kd). The system should be pretty reasonable under these changes.

Testing

Since a PID is inherently real-time and dependent on external systems, you can't really write code to test it quickly without decoupling it from the wall clock or having an external system with which to integrate it. In order to prove to myself that it basically works, I wrote a very simple simulation of a heater attached to a tank of water, and some tests to see what the PIDs behavior is when attached to the heater simulation. The heater, when turned on, heats up at a certain rate, and if its temperature is above the ambient temperature will lose a certain amount of heat to the environment over time.

The first test should demonstrate that the dummy heater is basically working.

test.py 0

The second test should demonstrate that the system can bring the water to a given temperature and hold it there.

test.py 1

Caveats

This is basically first draft code. Since the object is relatively straightforward there's hardly any need. If this bothers you, go ahead and add them, though it is not inconceivable that I would eventually get around to it myself. Nevertheless, it should be usable as-is without too much difficulty.

This code depends on pid.compute() being called at a predictable interval. It won't work as well under conditions which delay the call by too much, such as use under a non-real-time operating system which is trying to do too many things.

There are other PID features which could be added to make this even better. I'm not even sure I know what they are. If you do, feel free to write them and submit a pull request.

P.S. "jataruku" is a word in the Warlpiri language which according to the Australian Society for Indigenous Languages' Walpiri lexicon means "stubborn, bullheaded, hot-tempered."