4. Integration

Description

This guide explains how to control your car in real-time and dynamically update the PID parameters as the car runs on the track from inside a Jupyter notebook. To achieve that, we will configure Jupyter to start automatically at the boot, install ipywidgets which would allow us to interact in real-time with the Jupyter program, and configure the PWM to be controlled without sudo such that we can control the car speed and the servo angle within the notebook.

Important, we will use Jupyter notebook and not lab as we did before. Both programs have the same functionality, however, lab is under active development and it is difficult to find a version of widgets that will work with the current Jupyter lab version. Since widgets are an important part of this tutorial we will use Jupyter notebook instead of lab.

Configure Jupyter to start automatically at boot

In Linux, as soon as the kernel gains control over the CPU, it starts the init program that start all other services required by the operating system to function properly. Although, there are multiple programs that do the init functionality, the majority of the current Linux distributions, including Raspbian, are running systemd. Systemd is configured using text files that indicate what program to run and on what other services it depends. In our case, we will create a systemd service to start Jupyter notebook at the end of the boot process so you can work on your program without the need to ssh first and start the Jupyter notebook.

Step 1. Set Jupyter password

By default, Jupyter uses tokens to authenticate the users. When Jupyter starts, it generates a random sequence that is printed in the console, and asks you to copy-paste it into the browser such that it knows that you are the one who started the program. Since we will start it as a services, we would not have access to the sequence, as it will be printed into a log file. To make it work, we should enable password authentication in Jupyter which will disable the token authentication. To set the Jupyter password, first activate raceon environment by running workon raceon, then create the notebook configuration file where the encrypted password will be stored jupyter notebook --generate-config and set the password using jupyter notebook password that will ask for the password and update the configuration.

Step 2. Create a startup script that activates raceon environment and starts Jupyter server such that we can run it using systemd automatically at boot. In the home folder of user pi, use nano, a command line text editor, to create this script file. Type nano startjupyter.sh and paste the following content:

#!/bin/bash

source /home/pi/.virtualenvs/raceon/bin/activate

jupyter notebook --ip 0.0.0.0 --port 8888

Save the file by pressing Ctrl+O then Enter and exit using Ctrl+X.

Now, make the script executable using chmod +x startjupyter.sh and run it to check that the Jupyter notebook starts by typing ./startjupyter.sh in the terminal and then open in browser http://raspberrypi-ID:8888/ where ID is your pi number. Stop the script by pressing twice Ctrl+C

Step 3. Create the systemd configuration file that will instruct systemd to run the startjupyter.sh script as the pi user after Linux is ready to accepts user logins. Use nano to create the configuration file sudo nano /lib/systemd/system/raceon-jupyter.service and paste the following content:

[Unit]

Description=Service to start jupyter automatically

After=multi-user.target

[Service]

Type=idle

ExecStart=/home/pi/startjupyter.sh

User=pi

WorkingDirectory=/home/pi

[Install]

WantedBy=multi-user.target

Next, set the correct permissions for this file using sudo chmod 644 /lib/systemd/system/raceon-jupyter.service then tell the systemd to reload its configuration sudo systemctl daemon-reload and enable raceon-jupyter service to start automatically sudo systemctl enable raceon-jupyter.service

Before rebooting, check that the service starts without any errors using sudo systemctl start raceon-jupyter.service us and access again the notebook using the browser. In case there are errors, run sudo systemctl status raceon-jupyterlab.service that will print the console output of the service including the error message.

• Original post on piazza of this procedure is @38

Install ipywidgets

Ipywidgets is a python package the makes the python code running inside Jupyter interactive by providing user controls such as sliders, buttons, check boxes and other similar interface elements. To install ipywidgets, open in browser the main page of the notebook and select New->Terminal to open a terminal, and type pip install ipywidgets to install the widgets package using pip. After the installation is completed, create a new Jupyter notebook and paste the code below to test the new package:

from ipywidgets import interact

# Function that will be called for each slider movement

def f(x):

    print(x)

# Create a slider and set its default position to 10

interact(f, x=10)

The above code will create a slider below the cell and each slide move call the function f with x equal to the slide position. Later in this tutorial we will use the same concept to change the PID coefficients in real-time while the car is driving on the track.