Incrementally Smoothed Radio SLAM: A Factor Graph Approach to Opportunistic Radio Navigation

This paper describes a method for aiding an inertial navigation system (INS) with opportunistic radio measurements from cellular towers to maintain accurate positioning, navigation, and timing (PNT) in the event of a GPS outage. State-of-the-art approaches to opportunistic radio navigation apply the extended Kalman filter (EKF) to solve a radio-based simultaneous localization and mapping (SLAM) problem. This paper improves on these approaches by reformulating the existing EKF solution into an incremental nonlinear least-squares (NLS) solution using a factor graph. While the EKF has demonstrated sustained success in the general field of inertial navigation, it is poorly suited to SLAM problems due to its sensitivity to nonlinearities and accumulation of linearization errors. The proposed method circumvents these issues by performing iterative optimization over the entire trajectory. The resulting navigator demonstrates positioning error reduction of more than 30% over the state-of-the-art solution while maintaining a comparable computational burden.