The AlgoNav
Positioning Engine

Full multi-constellation, multi-frequency GNSS processing with automatic precise product retrieval. Self-adapting error models handle ionospheric, tropospheric and multipath effects without manual tuning — the engine determines optimal settings from the data itself.

Moving VRS — our key differentiator: Traditional post-processing software only supports stationary reference stations. For a maximum baseline distance of 20 km, that means managing dozens of VRS requests for a 300 km corridor — with handover issues, interpolation artifacts, and semi-automated workarounds. AlgoNav computes GNSS error terms directly at the vehicle's position, epoch by epoch. No approximation through a nearby virtual station. This yields better ambiguity resolution at longer baselines and eliminates the complexity of traditional VRS workflows entirely.

Sensors work together, not sequentially. GNSS, IMU, LiDAR, odometry and barometer inform each other in a tightly coupled Kalman filter — resulting in robustness that loosely coupled systems can't match. No static alignment phase required.

Why Tightly Coupled? In a loosely coupled system, each sensor computes its own solution independently before combining results. Tight coupling feeds raw observations from all sensors into a single estimator. The benefits are significant: better outlier detection even with only 2–3 visible satellites, more reliable quality estimates through full sensor redundancy, and maximum utilization of every available measurement.

Adaptive Residual-Based Tuning: The engine continuously analyzes filter residuals and automatically adjusts sensor noise models during processing. This adaptive mechanism ensures optimal weighting across varying signal environments.

Our engine goes beyond textbook forward-backward smoothing by applying Iterative Optimal Smoothing combined with robust outlier detection. This approach — only possible in post-processing — systematically improves trajectory quality in ways real-time systems fundamentally cannot achieve.

Highly efficient runtime. Efficient C++ core with CPU parallelization — no GPU required. Scale from single datasets to massive campaigns without specialized hardware.