Certified Hardware Configs¶
Community-contributed configurations that have been validated on real hardware. Each entry includes the specific sensor models, the deployment context, and a copy-paste ready YAML.
Certification levels¶
| Badge | Meaning |
|---|---|
| Field validated | Tested on a real robot outdoors with RTK or measured ground truth. Results are documented. |
| Community contributed | Contributed by a real deployer who ran it on their hardware. No RTK ground truth, but it runs without diverging. |
| Validation in progress | Configuration is correct per datasheet specs. A real user is integrating it and will update when validated. |
Registry¶
| Config | IMU | GPS | Platform | Status |
|---|---|---|---|---|
| Microstrain 3DM-GX3-45 + Novatel SPAN-CPT | Microstrain 3DM-GX3-45 | Novatel SPAN-CPT (~3m CEP) | Segway RMP campus rover | Field validated (12 sequences, 940 min, RTK ground truth) |
| Bosch BNO085 + u-blox ZED-F9P | Bosch BNO085 (9-axis) | u-blox ZED-F9P (standard GPS or RTK) | Differential drive outdoor | Community contributed (Oakland University IGVC robot, Agroecology Lab) |
| Xsens MTi-680G | Xsens MTi-680G (IMU + integrated GNSS) | Xsens MTi-680G GNSS (RTK capable) | Formula Student race car (Ackermann) | Validation in progress (UniNa Corse, FSAE Italy 2025) |
Contribute your config¶
If FusionCore is running on your robot, contribute your config. It gets credited to you, indexed by Google, and helps the next person with your hardware skip 3 hours of tuning.
What to submit:
A pull request to fusioncore_ros/config/ with:
- A YAML file named
<imu-model>_<platform>.yaml(e.g.,vectornav_vn200_husky.yaml) - A doc page at
docs/certified-configs/<imu-model>-<platform>.mdusing any existing page as a template - A one-line addition to the table above
Minimum for a "community contributed" entry: - You ran FusionCore on a real robot (not simulation) - The filter converged and tracked your position without diverging - You can state your sensor models and ROS distro
To reach "field validated": - Include a rosbag or trajectory plot alongside ground truth (RTK, measured loop, or GPS reference)
Open a pull request or start a discussion if you want help formatting the submission.
Why model-specific configs matter¶
The generic hardware configs (bno085_custom.yaml, clearpath_husky.yaml) give you correct starting points but say "datasheet-tuned, not yet field-validated." A certified config gives you noise values that were actually verified against real GPS error statistics on a real robot. The difference in imu.gyro_noise between a 3DM-GX3-45 and a BNO085 is 6x. Using the wrong value adds 20-40% to your ATE on long runs.