Talking Treebot

A mobile chatbot essentially used to make citizens interact with trees and foster their empathy for trees as organisms in our daily life.

⚠️ Update: We noticed that the multi-threading approach and playing back sounds on a Raspberry Pi Zero is a too intensive process, meaning that all the tasks will be performed in a very slow manner. Thus, we highly recommend using a Raspberry Pi with at least 4GB RAM. In our setup, we are working with a Rapsberry Pi v4 witht 4GB RAM.

Hardware

• Raspberry Pi Zero v1.1 (deprecated) / instead use: Raspi v4 with at least 4GB RAM
• 32GB micro SD card
• USB speaker box
• Bluetooth Lavalier microphone with USB receiver

Setup Raspberry Pi v4

The chatbot uses Rasbian OS (Bookworm) to save some RAM and energy. Please note that we use the latest Bookworm release in which Python v3.11.2 is pre-installed. We use Python 3.11.2 with pip as package manager and virtualenv to create isolate virtual environments.

1. Install Rasbian Bookworm OS by using the Raspberry Pi Imager to flash your SD card.

In case you use an older OS of Raspbian you could go with Python v9 (pre-installed).

Setup Environment

Enable I2C on Raspberry to make BME280 and button work: sudo raspi-config nonint do_i2c 0

System-level dependencies that cannot be installed with pip. Rather use aptas system-wide package manager to install the following: sudo apt update sudo apt install libasound2-dev sudo apt install portaudio19-dev sudo apt install mpg123 sudo apt install git

1. Install the same Python version you are using on your Raspberry Pi (ideally v3.11.2 any other version like v9 also works)
2. Pull this GitHub Repository by running git clone https://github.com/technologiestiftung/talking-treebot
3. Navigate into the folder cd talkding-treebot
4. Install venv and create a virtual environment: python3.11 -m venv treebot-env
5. Activate venv: source treebot-env/bin/activate
6. Install missing environment packages with pip install [the-missing-package] or simply run pip install -r requirements.txt

Configure Microphone and Speaker

Install the ALSA service utilities and follow steps below.

1. sudo alsactl init
2. list available playback devices (speaker) on RasPi aplay -l
3. list available capture devices (microphone) on RasPi arecord -l
4. Set both devices as default and edit sudo vim /etc/asound.conf

If your speaker is on card 1, device 0, and your microphone is on card 0, device 0, your /etc/asound.conf might look like this:

# Default speaker
pcm.!default {
    type hw
    card 2
    device 0
}

ctl.!default {
    type hw
    card 2
}

# Default microphone
pcm.!default {
    type hw
    card 3
    device 0
}

ctl.!default {
    type hw
    card 0
}

5. Restart the ALS service to apply the changes: sudo systemctl restart alsa

In order to increase the volume of your speaker e.g. to 80% you can run amixer set Master 80%

You should test your speaker and microphone configurations: Record 5sec: arecord -D hw:0,0 -f cd -t wav -d 5 -r 44100 -c 1 test-record.wav Play your recording: aplay test-recording.wav

Manually starting the Script

Note: running the script with sudo can cause other issues, such as conflicts with PulseAudio or the virtual environment. Try to avoid running main.py or button.py in sudo mode.

Instead give your user GPIO access to make button.py work: -Add your user (treebot) to the gpio group: sudo usermod -aG gpio treebot

• then reboot: sudo reboot
• check access (gpio should be listed): groups

Run: python main.py in order to start the conversation. I f you soldered a button to the Raspi, you can also run python button.py to run the main script whislt having more control over the interaction.

Automatically Startng the Script with System Services (preferred)

In order to have the script running automatically we are using two system services named button.service and treebot.service. Those should be place inside the following path: /home/treebot/.config/systemd/user/

After you've moved the files from this folder to the systemd folder the rapsberry Pi will start the main.py script automatically after sucesfully connected to a Wifi network. Afterwards, the button.py will be automatically started

treebot.service

[Unit]
Description=Treebot service
After=network.target

[Service]
ExecStart=/home/treebot/talking-treebot/treebot-env/bin/python -u /home/treebot/talking-treebot/main.py
WorkingDirectory=/home/treebot/talking-treebot/
#Environment="PATH=/home/treebot/talking-treebot/treebot-env/bin:$PATH"
Restart=always
StandardOutput=append:/home/treebot/treebot.log
StandardError=append:/home/treebot/treebot.err

[Install]
WantedBy=default.target

button.service

[Unit]
Description=Treebot button service
After=network.target
After=treebot.service

[Service]
ExecStart=/home/treebot/talking-treebot/treebot-env/bin/python /home/treebot/talking-treebot/button.py
WorkingDirectory=/home/treebot/talking-treebot/
#Environment="PATH=/home/treebot/talking-treebot/treebot-env/bin:$PATH"
Restart=always

[Install]
WantedBy=default.target

How to interact with my little Pi?

An older version of this project was using a headless OS (without screen, keyboard, mouse) which is why we needed to connect to the Raspberry via SSH.

1. run ping raspberrypi
2. copy the IP address and run ssh hostname@[copied-IP-adress] (e.g. 192.168.4.123)
3. connect with username and password

If you can't find the raspberry in your network, check if you laptop and the raspberry are connected to the same Wifi. To see all devices that are connected to your network run arp -a

Furthermore, we use magic wormhole to transfer files from the raspberry to our computer.

• Install on Raspberry Pi running sudo apt install magic-wormhole
• Install on MacOS running brew install magic-wormhole

In case you are using a Raspberry Pi with arm64 architecture (which is the case Raspi v4) you can improve your developer experience connect your Visual Studio Code IDE.