Raspberry Pi Motor Control Software

This is from one of my early robots, a single Raspberry Pi connected to an Adafruit Motor Shield. I wrote the Skidder C++ class to drive the vehicle with skid steering. See how it works...

TODO ~ replace image with one of the Raspberry Pi Controller

This software adds a Skid Steer motor control software to the Raspberry Pi Device Server device server. RPID adds motor controls accessible via both MQTT and REST APIs, as well as directly with the Skidder python library.

This project was to build a simple automous driving vehical using a single Raspberry Pi and a simple DC motor controller. The RPi acts as the macro controller making control decisions and managing communications.

The same RPi is also the Micro-controller, in other words on this Robot the RPi is both the Macro and Micro controllers, making the control decision as well as executing the control commands.

Controlling The Motors

This software is a component that plugs into the Raspberry Pi Device Server (RPID) allowing the motors to be controlled directly by the Skidder API, or via MQTT messages and/or HTTP REST requests. Oh, there is also a keyboard controller if you ssh into the vehicle.

TODO Place image 5-17 img-1 here

MQTT Skid Channel

The RPID MQTT server will subscribe to the /control/skid channel and respond to commands accordingly. Skid channel is simple, expecting two floating point numbers, representing the load (voltage) supplied to the motors.

% mosquitto_sub -t /control/skid 

Messages to control our skid steer vehical over MQTT will basically be two floating point values between -1 and +1 inclusive. The numbers represent the percentage of the load to let the motors have (throttle) and the +/- sign represents the direction the motor will turn.

/control/skid:l/r

REST Skid Endpoint

GET /skid/ PUT /skid/l/r

Get or Set the left and right throttles. The throttles values are represented by floating point numbers between -1 and 1 inclusive, representing the percentage of load, and direction that will be applied to the right and left throttles.

Adafruit RPi Motor Controller

For this Robot, we are using the wonderful Adafruit Raspberry Pi DC/Stepper motor controller. This is the one robot that we directly drive a motor controller from the Raspberry Pi. That aspect of this project differs than our other robots that are built with distributed controllers used by the other robots I have built.

RPi Macro Controler and Micro Controller

In other cases the Master Controller software runs run on a Raspberry Pi (or most any other debian linux host like a laptop or Jetson Nano) where as the Micro Controller typically runs on an embedded Arduino like device, possibly an esp8266, esp32 or stm32.

Localized Micro and Macro Controllers

In this software we will keep the same message based communication system, as well as REST support architecture as used on the other robots. In this case we will implement the Micro-controller requirements as part of the RPi DeviceService used by the Raspberry Pi Camera server.

That means we will be able to control the voltage sent to (throttle) the motors and hence control speed and direction the vehicle moves.

A Note about Performance and Communicaiton Overhead

In this case the RPi with minimal sensors, low speed motors and maybe a camera should be easily at home in a single multi-core RPi. This being the case, we could optimize communications by removing the messaging and translation layers from the MBR Architecture.

However, for the sake of a single architecture and minimized set of software tools and configurations, we will leave the communicaiton layers in place.

Consistency and maintainance are extreemly important when it comes to real-world production systems. But would we choose this over performance in a production system?

The answer depends on the demands of the application, in this case, I suspect the level of communicaiton overhead will be minimal, or trivial when compared to the Bandwidth and other resources available. That I believe removing the communicaitons layer would provide no noticible, production quality performance improvements. And at the cost of maintanance and complexity.

Of course speculation takes a deep backseat to real world numbers and careful testing. So, I reserve the right to be wrong!

Brushed DC Motors

We are using Brushed DC motors, about the cheapest possible motors one can buy. These particular motors accept 6 to 12v. Powering bots is particularly important, if you want to chase your tail on weird problems, work with under powered electronics for a while!

Powering The Vehicle

We can not share power with the Raspberry Pi as it does not have nearly enough power to run two motors, especially during heavy load. Therefore we must seek out additional power to be supplied independently to the Motor Controller.

Motor Control Software

The control software that rides on the raspberry pie will consist of, A copy of an empty MQTT broker, most likely mosquito. It will also consist of The control software described in (Control Software Design)[/notes/control-software-design] in the notes section of this site.

Speed and Steering

In this case we have a two wheel vehicle, with two motors. Our control Algorithm will employ skid steering, The simplest possible navigation mechanism we could develop.

Hence, The control algorithm will take its inputsTranslate them into skid steer during instructions,Which basically consists of varying the velocityOf the right and left group of wheels, In our case two wheels.

Outputs

Since the motors are controlled directly by Adafruit software, we will may create an individual process, separate from the control software just to handle the motor controls.

Bill of Materials

  • Raspberry Pi V3+
  • Adafruit Motorsheild
  • Extra battery pack for motors
  • 2 brushed DC motors
  • 2 wheeled cheap balancing disk for body
  • Application specific sensors