FusionCore brings a new ROS 2 sensor fusion SDK that combines IMU, wheel encoders, and GPS data into reliable position estimates for mobile robots. This package addresses critical gaps in existing solutions like robot_localization and fuse, offering native ECEF GPS fusion, IMU bias estimation, and adaptive noise covariance without manual tuning.

Benchmark results show FusionCore outperforms alternatives in 5 of 6 test sequences using the NCLT dataset, with position errors ranging from 4.2m to 28.7m versus 5.8m to 23.4m for competitors. The UKF filter implementation handles challenging conditions like GPS degradation better than standard approaches while maintaining robust performance across diverse environments.

Released under Apache 2.0 license, FusionCore installs as a monorepo with four independent packages and works with ROS 2 Jazzy. Its lifecycle node approach ensures proper initialization before processing data, while comprehensive testing capabilities allow developers to verify functionality without physical hardware, making it accessible for both prototyping and production deployment.