Real-time loosely coupled GNSS and IMU integration via Factor Graph Optimization

A team of researchers including Radu-Andrei Cioaca and Florin Stoican has published a breakthrough paper on arXiv detailing a new real-time method for fusing Global Navigation Satellite System (GNSS) and Inertial Measurement Unit (IMU) data using Factor Graph Optimization (FGO). Their work addresses a critical bottleneck in autonomous navigation: maintaining reliable positioning in GPS-denied environments like urban canyons where satellite signals are blocked. Traditional FGO methods are computationally intensive and typically used for post-processing, but this new 'loosely coupled' architecture is designed for live operation, prioritizing constant service availability over maximum possible accuracy.

The technical innovation lies in the real-time implementation of the FGO framework, which was validated using the challenging UrbanNav-HK-MediumUrban-1 dataset. The results show the system achieves 100% service availability, meaning it never loses track of the vehicle's position, a crucial requirement for safety-critical autonomous systems. However, the paper provides a transparent analysis of the inherent trade-off: this real-time capability comes at the cost of reduced positioning accuracy compared to slower, batch-processing FGO methods. This research provides a practical pathway for deploying robust, always-on localization in self-driving cars, delivery robots, and drones operating in complex city environments.