The MQTT message bus is a simple but critical element of our MBR software design. MQTT provides what is known as a publish/suscribe or pubsub communications model. This model of machine to machine communication has some attractive benefits for realtime driving algorithms that we will discuss in this post.
Pubsub message bus and realtime communications
The pubsub communication model boasts some properties that are attractive to a real time application. Specifically, messages can be sent and delivered efficiently with little delay, provided none of the resources are over burdened.
Decoupling the Controller from Control Object
Traditionally, the software, and most likely the hardware intrinsicly ties a particular device with a specific controller. For example, a Joystick or Keyboard. Replacing, say the keyboard with Joystick control requires new software to be added to the controlled item, as well as created for the controller.
It also opens up opportunity for logging, learning, testing and replays!
Keyboards, programs or canned messages can be formed to sned the same communication.
Talk to The Bus
All entities talk to the bus, and read from the bus. No two components speak directly to one another. This allows an incredible amount of flexibility, but it does add a level of abstraction and compexity.
Adding a message bus requires us to adjust from a tight, input -> output coupling to a decoupled intention based messaging system.
For example, a Joystick can publish rapid stick movement / positioning informaiton to an MQTT bus such that subscribers like the Control software, a logger or debugger as well.
A temprature sensor may send out periodic temprature. Distance, light, heat and other sensor can periodical advertise to appropriate channels.
Funcation Call, REST vs. Pub/Sub
[Todo] put a link here to a page that compares communication models: such as Routing Protocols, Web/REST servers, RealTime sockets, Real-time Websockets, real time messaging.
MBR and the Pubsub Commuication Model
For our application, we will be controlling a set of DC motors through a micro-controller (i.e. Arduino with shield) attached via some serial/RF/BT/??? mechanism.
For this project we will focus on the MQTT protocol and we will specifically use Mosquitto the MQTT broker.
useful for many types of computer communication, including real time control software, like we use to drive our mobile vehicle. This article will go into detail on our controll system software and the various components required to make it happen!
The MQTT message server a general purpose as a systems message bus our project components can use to read and write data in near realtime. This data is used for a number of purposes, to advertise changes to control inputs and sensors, publish control messages to motor controllors and by the dashboard to display our realtime sensor data. This describes how our MQTT bus is being used and how we got it all hooked up!
MQTT Bus Uses
- Advertise Joystick inputs for vehicle control
- Advertise Sensor readings for sensor networks
- Consume vehical controls by the motor controllers
- Consume all sorts of data for the Dashboard and realtime graphs
Mosquitto MQTT Broker is running on a system wide Raspberry Pi. All sensor and control inputs are written to the appropriate topics on the MQTT bus.
In the future we’ll want to run an MQTT bus on vehical incase we loose contact with our control station.
- draw an image of the MQTT communication
- document the MQTT installation process for broker
- document otto mqtt client to output process
- show some code examples.
date: 2019-03-04 description: > This project has use create a micro service that provides access to the Raspberry Pi Camera and Adafruit Motor Controller. So I built a service that can be controlled via Web App, REST API or MQTT message bus. categories:
- micro service tags:
- pi camera
- motorkit todo:
- block diagrams
- link to github report image: https://mobilerobot.sfo2.cdn.digitaloceanspaces.com/display-and-motors.jpg resources: github: http://github.com/mobilerobot-io/rpid
MQTT works with three components: data publisher, subscriber and a broker. This model is commonly refered to as the pub/sub model.
The Collection Stations (CS) publish data to the MQTT broker. The broker then forwards the data to all of the subscribers, which includes the IoT Hub.
MQTT uses topics are simply strings that have a syntax very similar to a filesystem path used by a computers operating system. The elements that make up a topic are sepearated by the ‘/’ characture. The words between the slashes give the topic structure and semantics.
MQTT clients can use wildcards when subscribing to topics allowing the subscriber to recieve data from multiple unknown publishers.
This application provides access and control to various Raspberry devices like the camera, GPIO pins, or a possible MotorShield if needed.
Access is provided via a Webapp and REST API thanks to Flask. Control can also be done by sending the appropriate messages to specific MQTT message buses.
I built a Raspberry Pi mobile robot [r3] with just a Raspberry Pi and the Adafruit RPi MotorController, if the MotorController is present the corresponding Adafruit motor controller library will be installed and used.
If the Motorshield is not present, the motorcontroller library will not be loaded, or a Fake (do nothing) library will be loaded on devices that do not including the Adafruit MC, which includes non-Raspberry Pi platforms.
TODO: put a link to installing the Motorshield library, and a quick cheat.
Motors and Skid Steering
The ‘motors’ provides the library
Skidder that provides a Skid
Steering steering control, that uses the Adafruit MotorKit library to
control the throttles of the motors being used.
For a brief overview of Skid Steering refer to this article
That article will describe this project within the bigger Mobile Robot Architecture (MRA).
Pi Cameras and Video Control
Another big hunk of the device server is controlling the Video camera. We have the ability to start and start recording video, or take snapshots.
The videos and images can be transfered to a remote host if a network is available, they can be saved on vehicle if no network is available until one is.
The camera is also able to stream video over a network, if present, for real time consumption of the video as it is being filmed.
Video is Consumed by Humans and Computers
The video is made accessible to Humans for and Computers alike by multicasting the video over an IP network with an IP multicast address.
Read more about IP Multicast here
Refer to the PiCamera documentation here: https://picamera.readthedocs.io/en/release-1.13/.
The software can be found in github repository.
And how to write and operate a flask based server for your project.
TODO ~ Turn streaming On and Off
The next task is to turn video streaming On and Off, this may require duping and modifying somewhat a python script …