Hi everyone,
I’ve been building FusionCore over the past few weeks… It’s a ROS 2 Jazzy sensor fusion package designed to replace robot_localization. The filter is working correctly in simulation, and I’m now looking for people to test it on real hardware.
What it does is that it essentially combines IMU, wheel encoders, and GPS into one reliable position estimate at 100Hz. UKF-based, ECEF GPS natively, automatic IMU bias estimation, Mahalanobis outlier rejection, adaptive noise covariance. One YAML config file, no manual covariance tuning.
I’m specifically looking for:
- Anyone who has hit the robot_localization deprecation wall and hasn’t found a clean path forward
If you’re willing to share your setup, I will personally help you get it running…. and I’ll build around your specific configuration. Your robot, your sensors, your use case becomes the thing I optimize for next. Whatever doesn’t work on your hardware is what I fix first.
Open a GitHub issue, reply here, or DM me directly.
GitHub: https://github.com/manankharwar/fusioncore
2 Likes
Hi Mahan,
I’d love to test your package.
Our robot is mecanum wheel driven robot. We want to fuse the wheel odometry with our IMU. We’ve had our difficulties getting this to work with robot_localization, as on the uneven floor sometimes a wheel looses traction. Do you think your implementation can cope with the wheel odometry outliers when the robot drives over an uneven spot?
Hi, thanks for reaching out…. this is exactly the kind of use case I want to test on real hardware.
Short answer: yes, wheel odometry outliers from slip are handled. FusionCore uses Mahalanobis gating on encoder measurements…. if a wheel loses traction and the velocity reading becomes statistically implausible given the current state, it gets rejected and the filter runs on IMU alone until odometry recovers.
One honest caveat: FusionCore currently uses a differential drive motion model. Mecanum kinematics aren’t explicitly modeled yet…. lateral velocity won’t be fused correctly. That’s the next thing I’d build if you’re willing to test it.
What IMU are you using and how are you publishing odometry? I’ll help you get it running and whatever breaks on your hardware is what I fix first. Thanks again… We can connect/DM as well.
1 Like
Sounds good, I’ll send you a DM
1 Like
My user here is too new, I can’t send you a DM. I’ve sent you a message to your university email instead
1 Like
Great! I’ve replied back Marc… Let me know if you have difficulties accessing it.
1 Like