CommaCarSegments: 3148 hours of raw CAN bus data from 230 different car models

commaCarSegments is a dataset of raw CAN bus data recorded from our fleet of openpilot users driving over 300 different production vehicles, around the USA and rest of the world.

Structure

. ├── segments/ # drectory containing all data │ ├── <device_id>/ # unique device ID directory │ │ └── <route_id>/ # unique route (aka drive directory) │ │ └── <segment>/ # unique segment index inside of the route │ │ └── rlog.zst # actual data └── database.json # manifest for the dataset

The rlog.zst is a zstandard-compressed archive of openpilot cereal messages that can be read with openpilot's LogReader.

Using this dataset

Here's a simple example that shows the CAN addresses with the length of their messages on a Chevy Bolt:

In [1]: from openpilot.tools.lib.logreader import LogReader In [2]: lr = LogReader("da4f7891ae804737/00000341--589a82bd48/20/s") In [3]: {c.address: len(c.dat) for c in lr.first("can")} Out[3]: {190: 7, 201: 8, 193: 8, 197: 8, 209: 7, 840: 5, 842: 5, 485: 8, 241: 6, 211: 3}

Also see:

• opendbc: our project for interfacing with the car's CAN bus
• cabana: a beautiful tool for exploring CAN data (try ./cabana --demo)
• Example notebooks using this dataset for large-scale validation of openpilot changes

In the dataset

Totals

• 230 different platforms
• 188883 segments
• 3148 hours

We're hiring!

If you have fun playing with this dataset, consider a job working on this stuff here at comma.