GNSS-Inertial State Initialization Using Inter-Epoch Baseline Residuals
Javier Civera, Samuel Cerezo
AI summary
Problem
Initializing GNSS-inertial systems with limited early measurements often produces poor estimates that converge to local minima, while naive early fusion of all data causes inconsistent estimates and divergence.
Approach
The method employs a two-stage initialization that postpones global GNSS position constraints until an observability-based criterion is met. Initially, it leverages inter-epoch baseline vector residuals between consecutive GNSS fixes to mitigate inertial drift and ensure geometric consistency.
Key results
- Observability-based triggering criterion using Hessian singular values to activate global constraints
- Inter-epoch baseline residual formulation that mitigates inertial drift without high-rate GNSS
- Consistent outperformance of naive early-fusion strategies across EuRoC, GVINS, and MARS-LVIG datasets
- Robust navigation bootstrap for UAVs and mobile robots without external heading or gravity priors
Why it matters
Provides a reliable, data-driven initialization strategy that prevents divergence and improves navigation accuracy for UAVs and mobile robots operating with GNSS-inertial sensors.
Abstract
Initializing the state of a sensorized platform can be challenging, as a limited set of measurements often provide low-informative constraints that are in addition highly non- linear. This may lead to poor initial estimates that may converge to local minima during subsequent non-linear optimization. We propose an adaptive GNSS–inertial initialization strategy that delays the incorporation of global GNSS constraints until they become sufficiently informative. In the initial stage, our method leverages inter-epoch baseline vector residuals between consecutive GNSS fixes to mitigate inertial drift. To determine when to activate global constraints, we introduce a general criterion based on the evolution of the Hessian matrix’s singular values, effectively quantifying system observability. Experiments on EuRoC, GVINS and MARS-LVIG datasets show that our approach consistently outperforms the naive strategy of fusing all measurements from the outset, yielding more accurate and robust initializations.