MaRS: Truly modular multi-sensor fusion framework now available
If you are interested in modular sensor-fusion with minimal State-Representation despite a high number of sensors and their calibration states, then you might also be interested in our first release of the MaRS Framework.
We are happy to announce that our in-house developed state-estimation framework “MaRS” is now available on github. You can add any number of sensors to a setup and use their information efficiently. The core software is provided as a stand-alone C++ library, ready for the integration in any middle-ware tool. We also provide a ROS wrapper for this library with example setups to get your project off the ground quickly.
Feel free to visit the MaRS GitHub pages.
- Stand-Alone C++ Library GitHub - aau-cns/mars_lib: MaRS: A Modular and Robust Sensor-Fusion Framework
- MaRS ROS Wrapper GitHub - aau-cns/mars_ros: A ROS wrapper for the MaRS Library
Some of the Highlights
MaRS Framework (Stand-Alone C++ Library)
- Truly modular decoupling of sensor-states from the core navigation-states
- Generalized covariance segmentation for plug and play state and covariance blocks
- Minimal state-representation at any point in time
- Integration and removal of sensor modules during runtime
- Out of sequence sensor measurement handling
- Developed for computationally constrained platforms
- Efficient handling of asynchronous and multi-rate sensor information
- Separation between simple user interaction and the complexity of information handling
- Google tests and Docker test environment
ROS Wrapper
- Ready to use ROS nodes for common setups (position, pose, and GNSS sensors with IMU)
- Predefined sensor update modules (plug and play)
- Predefined RQT views
- Docker test environment
More Information
The RA-L paper: MaRS: A Modular and Robust Sensor-Fusion Framework | IEEE Journals & Magazine | IEEE Xplore
Method animation:
Real-world demo video:
Looking forward to your feedback and videos of your MaRS powered autonomous robots,
Christian Brommer, Roland Jung, Jan Steinbrener and Stephan Weiss