Control a Robot Dog with a Custom Viam Base Component

The base component type is useful for controlling mobile robots because it gives users intuitive steering controls to use in code as well as from the Viam app remote control interface.

Viam natively supports a wheeled base model, but if you have a quadruped or other form of base that requires a different underlying implementation, you can create a custom component as a modular resource.

This tutorial demonstrates how to add a custom base using this robot dog kit and its open source code as an example.

By the end of the tutorial, you will be able to drive this dog around using the Viam base methods: MoveStraight, Spin, SetPower, SetVelocity, and Stop. You will also be able to use the CONTROL tab in the Viam app to remotely drive the dog around using your keyboard while viewing the camera feed. You’ll learn to implement a custom component type in Viam, and you’ll be equipped to implement other sorts of custom components in the future for whatever robots you dream up.

Code used in this tutorial

Hardware requirements

Raspberry Pi setup

Freenove documentation includes Raspberry Pi setup instructions but we recommend the following steps to make sure the Pi is set up for this tutorial:

Follow the steps in our Raspberry Pi Setup Guide to install Raspberry Pi OS.

Add a new machine in the Viam app. Then follow the setup instructions to install viam-server on the computer you’re using for your project and connect to the Viam app. Wait until your machine has successfully connected.

  1. SSH into the Pi to complete the following steps.

  2. Install pip and then git:

    sudo apt install pip
    
    sudo apt install git
    
  3. Navigate to the directory on the Pi where you’d like to install the Freenove robot dog code (for example /home/fido/). Get the code by running the following command:

    git clone https://github.com/Freenove/Freenove_Robot_Dog_Kit_for_Raspberry_Pi
    
  4. Check which version of Python you have installed on the Pi:

    python --version
    

    If it isn’t Python 3.8 or later, install an updated version of Python and double-check that you’re running the latest Raspberry Pi OS.

  5. Install the Viam Python SDK:

    pip install viam-sdk
    
  6. Enable I2C per the instructions in the Raspberry Pi Setup Guide.

  7. Alter the I2C baud rate according to Chapter 1, Step 2 in the Freenove instructions (page 40 as of January 24, 2023).

  8. Install smbus so that the servo code works:

    sudo apt-get install python3-smbus
    
  9. Follow Chapter 1, Step 3 (page 42 as of January 24, 2023) of the Freenove tutorial to complete the software installation:

    cd /home/fido/Freenove_Robot_Dog_Kit_for_Raspberry_Pi/Code
    sudo python setup.py
    
  10. Restart the Raspberry Pi:

    sudo reboot
    

Hardware setup

Follow the Freenove hardware assembly instructions, including servo setup (all of Chapters 2 and 3, and the section of Chapter 4 about calibration) before proceeding.

Check connection between the robot dog server and the midlayer

Before proceeding with the custom component implementation, follow the instructions in this section to test the connection between the Freenove server running on your robot dog and the code running on your development machine (laptop or desktop). This way, you can isolate any client-server connection problems if they exist.

Create a connection test file

In a convenient directory on your development machine, create a Python file and open it in your favorite IDE. We named ours dog_test.py and opened it in Visual Studio Code.

Paste the following code snippet into the file you created:

# dog_test.py is for testing the connection
import socket
import time

s = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
s.connect(("PASTE DOG IP ADDRESS HERE", 5001))

cmd = "CMD_MOVE_FORWARD#15"
s.send(cmd.encode("utf-8"))
time.sleep(7)
cmd = "CMD_MOVE_STOP"
s.send(cmd.encode("utf-8"))
cmd = "CMD_RELAX"
s.send(cmd.encode("utf-8"))

Save the file.

Find IP address

Go to the machine page for your robot dog that you created when installing viam-server on the Pi.

Open the part status dropdown in the top left of the page. If your machine is connected to the app, this should say Live. The IP address of the robot dog Pi is displayed under IPs. Click the copy icon to copy the IP address to your clipboard. Inside dog_test.py, replace PASTE DOG IP ADDRESS HERE with the copied IP. Save the file.

Test the connection

Place the robot dog on a flat, open surface where it has room to take a few steps.

Now you are ready to run the connection test file.

Open two terminal windows on your development machine: one for the robot dog Pi and one for the development machine.

In one terminal, SSH into the Pi using the username and hostname you set up when imaging the Pi, for example:

ssh fido@robotdog.local

Navigate to the /home/fido/Freenove_Robot_Dog_Kit_for_Raspberry_Pi/Code/Server directory.

Start the robot dog server by running:

sudo python main.py -tn

In the other terminal window, navigate to the directory on your development machine where you saved dog_test.py. Run the connection test file with the following command:

python dog_test.py

If the dog walks forward for a few seconds and then stops, the connection works and you can successfully send commands to the robot dog server. Continue on to implement the base.

If the robot dog did not respond, double check your IP address, make sure the robot dog server is still running, and make sure the robot has adequate power. You can also try turning the robot off and on again, and then retrying the process.

Implement the custom base code

Now that the Freenove server is set up, you will follow the process for creating modular resources. You will use this module creation tool to simplify the process and generate the necessary stub files. Then, you will edit them as necessary to define how each base API method interacts with your robot dog.

Create the module files

First, from the Raspberry Pi terminal, create a directory inside the home directory to hold your custom code files:

mkdir robotdog

Next, install the module creation tool. Run yo viam-module and follow the prompts to create the files for your custom base. As you follow the prompts, it’s a good idea to use the same names we used so that your code matches the example code:

  • Model name: robotdog
  • Module namespace: viamlabs
  • Module repo-name: base
  • Language: python
  • API triplet: rdk:component:base
  • Existing API? Yes

Look in your robotdog directory and you’ll see the tool has created a requirements file and an executable. Look inside the src subdirectory which now contains __init__.py, main.py, and robotdog.py. robotdog.py is the file that defines the behavior of your custom base. This is the file you will modify in the next steps.

Connect the module to the Freenove server

When you send a command to the robot using the Viam base API, you need a way to pass the corresponding command to the Freenove dog server. In your code, establish a socket and then create a send_data helper method to send the command from viam-server to the Freenove server.

Start by importing socket:

import socket

Then add ip_address and port attributes to your custom base so you can configure the correct connection details in your machine’s config later. Modify the validate and reconfigure methods to use these attributes, and define the send_data method. The code below shows what the top of your class definition should look like.

class robotdog(Base, Reconfigurable):
    MODEL: ClassVar[Model] = Model(ModelFamily("viamlabs", "base"), "robotdog")

    # Class parameters
    ip_address: str
    port: int

    # Constructor
    @classmethod
    def new(cls,
            config: ComponentConfig,
            dependencies: Mapping[ResourceName, ResourceBase]) -> Self:
        my_class = cls(config.name)
        my_class.reconfigure(config, dependencies)
        return my_class

    # Validates JSON Configuration
    @classmethod
    def validate(cls, config: ComponentConfig):
        # Here we validate config, ensuring that an IP address was provided
        ip_address = config.attributes.fields["ip_address"].string_value
        if ip_address == "":
            raise ValueError("No IP address provided")

        port = config.attributes.fields["port"].number_value
        # Per the Freenove code, 5001 is for sending/receiving instructions.
        # Port 8001 is used for video.
        if port == "":
            port = 5001
        return

    # Define a way to send commands to the robot dog server
    def send_data(self, data):
        try:
            self.client_socket.send(data.encode("utf-8"))
        except Exception as e:
            LOGGER.error(e)

    # Handles attribute reconfiguration
    def reconfigure(self,
                    config: ComponentConfig,
                    dependencies: Mapping[ResourceName, ResourceBase]):
        # Here we initialize the resource instance
        ip_address = config.attributes.fields["ip_address"].string_value
        port = config.attributes.fields["port"].number_value
        self.client_socket = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
        self.client_socket.connect((ip_address, port))
        return

Define the custom base interface

To create a custom base model, you need a script that defines what each base component method (for example set_power) makes the robot dog do.

Open your newly created robotdog.py file. It contains stubs of all the base API methods, but you need to modify these method definitions to actually send commands to the robot dog.

Take a look at robotdog.py.

It defines each method by specifying which corresponding commands to send to the Freenove dog server when the method is called. For example, the stop method sends a command (CMD_MOVE_STOP#8) to the robot dog server to stop the dog from moving:

async def stop(self, extra: Optional[Dict[str, Any]] = None, **kwargs):
    self.is_stopped = True

    command = "CMD_MOVE_STOP#8\n"
    self.send_data(command)

Copy and paste that code into your robotdog.py file. Feel free to tweak the specific contents of each of the base method definitions to do things like make the dog move faster. Don’t forget to save.

Make your module executable

Now that you defined the methods for the custom component, you need to set up an executable file to run your custom component module. You can find more information in the relevant section of the modular resource documentation. Since the command line tool already created a run.sh for you, all you need to do is make that shell script executable by running this command from your robotdog directory:

sudo chmod +x run.sh

Configure the module on your robot

You need to tell your robot how to access the module you created.

On the Viam app, go to your machine’s CONFIGURE tab. Click the + (Create) button next to your main part in the left-hand menu and select Local module, then Local module. Name your module my-custom-base. Enter the path (for example, /home/fido/robotdog/run.sh) to your module’s executable file in the Executable path field. Click Save at the top right of the page to save your config.

Screenshot of the Viam app Config tab with the Modules subtab open, showing my-custom-base configured.

Configure the components

Now that the custom base code is set up, you need to configure all your hardware components.

Configure the camera

Configure the ribbon camera on the dog as a webcam following our webcam documentation.

Click Save.

Configure the base

Now, add the local base component from the local module. Click the + (Create) button next to your main part in the left-hand menu and select Local module, then Local component. Select base as the Type. Name your component quadruped. For the colon-delimited triplet, enter viamlabs:base:robotdog.

Then, in the configuration panel that appears for the local base, copy and paste the follow attributes:

{
  "ip_address": "<HOSTNAME>.local",
  "port": 5001
}

Edit the ip_address attribute to match your machine’s hostname, replacing <HOSTNAME> with your Pi’s hostname (for example, "ip_address": "robotdog.local"). If this doesn’t work, you can instead try using the IP address of the machine where the module is running, for example, "ip_address": "10.0.0.123".

If you are using a port other than 5001, edit the port attribute. 5001 is the default port for sending and receiving instructions to and from the Freenove server.

Click Save.

Your local component configuration should look similar to the following:

CONFIGURE tab, showing the attributes editor of the quadruped local base component.

Start the Freenove server

To operate the dog, you need to start the Freenove robot dog server (which you saved as /home/fido/Freenove_Robot_Dog_Kit_for_Raspberry_Pi/Code/Server/main.py).

Configure a process to automatically start the server on boot so it is ready to receive commands from viam-server.

Click the + (Create) button next to your main part in the left-hand menu and select Process.

In the process configuration panel that appears, click on the name in the top left corner and rename it to freenove. Fill out the configuration panel as follows:

  • Executable: The path to Python on your Pi, for example /usr/bin/python. If you’re not sure what this path should be, run which python on your Pi command line to find it.
  • Arguments: Type in main.py and click Add argument. Then type -tn and click Add argument again. This flag starts the Freenove server without launching Freenove’s GUI (which you don’t need for this use case).
  • Working directory: /home/fido/Freenove_Robot_Dog_Kit_for_Raspberry_Pi/Code/Server (changing “/home/fido” to the name of your home directory where you downloaded the Freenove code).
  • Logging: Toggle to the on position so you can view logs for this server process.
  • Execute once: Leave this off so that the Freenove server will continue to run, and will restart if it crashes.

Click Save.

Screenshot of the Processes subtab of the Config tab, showing a process configured as detailed above.

Click to see what the processes config looks like in JSON mode.

Driving the robot from the Viam app

Navigate to the CONTROL tab.

Click the quadruped component panel to expand it and reveal the controls.

Screenshot of the CONTROL tab with the custom base card expanded to reveal arrow control buttons.

  1. Enable the camera stream from the Select Cameras dropdown.
  2. Toggle the Keyboard Disabled switch to Keyboard Enabled to use the WASD keys on your keyboard.
  3. Use the W, A, S and D buttons to make the robot walk forward, turn left, walk backward or turn right.

Troubleshooting

  • If your servos aren’t moving as expected or at all, try turning the whole robot off for a while to let them cool down.

  • Make sure the machine’s batteries have adequate charge. If you have otherwise unexplained connection errors, try powering things off and charging the batteries for a while before attempting to SSH to the Pi again.

  • If certain sensors or servos aren’t being found by the software, turn off the robot and make sure all wires are fully connected before turning it back on.

  • If you want to send commands directly to the dog server instead of running robotdog.py (which may be helpful for debugging specific commands, especially if you’re adding your own functionality and need to calibrate servo speeds/positions) you can do the following:

    1. Install Netcat if it isn’t already installed:

      sudo apt install netcat
      
    2. Connect directly to the robot dog by running the following command (replacing with the correct IP address, for example nc 10.0.0.123) from the command line while SSHed into the Pi:

      nc <DOG IP ADDRESS> 5001
      
    3. You can now type commands (list of available commands) and hit enter to send them to the Freenove robot dog server. For example:

      CMD_TURN_LEFT#30
      CMD_STOP
      

Next steps

In this tutorial you learned how to implement a custom component model and control it using the Viam app. You learned about configuring modules and processes. You drove the robot dog around using the Viam CONTROL tab.

To add more functionality, try using the generic do_command method to add different behaviors to your robot dog. You could also use the Viam vision service with the robot dog’s camera component. For example, you could write code to tell the robot dog to move towards a colored target or to follow a colored line, similarly to how these tasks are done with wheeled bases in the tutorials linked here.

Have questions, or want to meet other people working on robots? Join our Community Discord.

If you notice any issues with the documentation, feel free to file an issue or edit this file.