Robin is an echolocation headset; it is a research platform, and a prototype mobility aid for blind people. It is the spiritual successor to the Sonic Eye (Sohl-Dickstein et al., 2015). This repository and this README focus on the code for the device; schematics for the hardware are in other repos in this Github organization. But for the curious, the device currently looks like this:
The code is written in Python and is designed to run on a Raspberry Pi.
You can skip this section if you already have a working device.
You'll need:
- A Raspberry Pi 4
- A MicroSD card (16 GB should be plenty)
- The HifiBerry DAC+ ADC Pro
Grab the Raspberry Pi Imager and pop in your SD card. You'll want the 64-bit version of Raspbian for your Pi. Make sure you choose the Edit Settings option before proceeding. You'll want to change the following:
- The hostname: you'll use this to access the Pi on the local network. We commonly call it
robin
, but if there's an existing device with that name floating around, you'll want to call it something else. - A username and password. We usually use the username
robin
. - The Wi-Fi SSID, password, and locale that you want the Pi to connect to when it boots
- The local timezone and keyboard layout
- Make sure to enable SSH from the Services tab.
The installer will install Raspbian on the SD card. Insert it into the Pi and turn it on. To connect for the first time, use:
ssh (username)@(hostname).local
Above, we chose robin
for both the username
and hostname
, so we connect using ssh [email protected]
. You will be prompted for your password. You'll probably want to follow these instructions so you can log in without a password in the future.
First, download OS-level libraries that Robin depends on. In particular, you'll need Poetry to manage our Python code, and pipx to install Poetry:
sudo apt-get install pipx libasound2-dev npm
pipx ensurepath
pipx install poetry
You will need to open a new shell (for instance by reconnecting via SSH) before poetry
becomes available. Then you can install this repo:
- Use
git clone https://github.com/robin-labs/robin.git
if you know you won't need to push your changes back to this repo - Use
[email protected]:robin-labs/robin.git
if you have permission to commit to this repo. Follow these instructions to add an SSH key to the Pi (not to your own machine) so you can push code directly from the Pi.
Now run a few install scripts:
# Enter the repo
cd robin
# Install Python dependencies
poetry install
# Install our stuff to the OS: device overlay, systemd service, etc.
poetry run rpi-install
If poetry install
hangs, try adding this variable to your shell:
export PYTHON_KEYRING_BACKEND=keyring.backends.null.Keyring
Now run sudo reboot
to restart your Pi. Then, if you run arecord -l
, you should see the HifiBerry sound card registered as an audio device:
**** List of CAPTURE Hardware Devices ****
card 2: sndrpihifiberry [snd_rpi_hifiberry_dacplusadcpro], device 0: HiFiBerry DAC+ADC Pro HiFi multicodec-0 [HiFiBerry DAC+ADC Pro HiFi multicodec-0]
Subdevices: 1/1
Subdevice #0: subdevice #0
Now you should be able to start the Robin software:
cd robin
poetry run start
Recordings will be saved to the ~/recordings
directory.
After you sudo reboot
, the device should work without further intervention from the command line.
The installation script registers a systemd
service called robin.service
. It will automatically start Robin on boot (if the headset is connected and powered on, you'll hear three clicks from the emitters when it comes online). You can view the output using journalctl
:
journalctl -u robin -f
You can use the start
, stop
, restart
, enable
, and disable
commands from systemctl
to manage the service, e.g.
sudo systemctl restart robin
Robin ships with a helpful web interface for debugging and testing (see the batcave_server
directory). By default, the Robin service will serve the remote over the local network. Assuming you're connected to the same Wi-Fi network as Robin, you can visit http://robin.local:8000
to visit the remote (assuming the hostname is robin
— you may have chosen something else). It has three tabs:
Emit
is a big button that will emit the currently configured pulse and stream it through your browserConfig
provides a frontend toconfig.json
(see the section below) and lets you control many important settings, particularly the output pulseLogs
shows debug logs streaming from the device.
Many important settings are specified in config.json
at the root of the repository. It is not checked into source, but the poetry run install-rpi
command should install a basic config.json
for you to use. The available settings are specified by Pydantic models in robin/config.py
, and these are mirrored in config.schema.json
which should give you good autocomplete support in an IDE like VSCode:
This lets you control settings like:
- The currently emitted pulse
- A bluetooth remote to connect to, and a mapping of
remote_keys
to pulses - Key echolocation settings, for instance the
slowdown
, and gain calibration options for the microphones - Which recordings are saved from each pulse.
The setup scripts install a Git pre-commit
hook which runs the Ruff linter and formatter for Python code before allowing commits. To format the code you can use:
poetry run ruff format
You can pass the -n
flag to git commit
if you want to skip this step.