CBL - Campus del Baix Llobregat

Projecte llegit

Títol: Uso de modelos digitales de elevación para las estimaciones del ángulo de máscara basadas en el terreno en algoritmos de predicción RAIM

Estudiants que han llegit aquest projecte:

Director/a: RAMONJOAN ESCOBAR, ÀLEX

Departament: FIS

Títol: Uso de modelos digitales de elevación para las estimaciones del ángulo de máscara basadas en el terreno en algoritmos de predicción RAIM

Data inici oferta: 19-01-2026     Data finalització oferta: 19-04-2026


Estudis d'assignació del projecte:
    GR ENG SIST AEROESP
Tipus: Individual
 
Lloc de realització: Fora UPC    
 
        Supervisor/a extern: Daniel Garcia Garcia
        Institució/Empresa: Pildo Consulting SL
        Titulació del Director/a: Enginyer de Telecomunicacions
 
Paraules clau:
Gnss, gps, raim, predicción, ángulo, elevación, terreno, modelo
 
Descripció del contingut i pla d'activitats:
The aim of this project is to investigate the potential use of freely available Digital Elevation Models (DEMs) to improve the prediction of RAIM availability outages at low altitudes by accounting for the surrounding terrain. Terrain obstructions may block the visibility of GPS satellites, thereby degrading the satellite geometry and impacting RAIM performance.

The first phase of the project will focus on researching and evaluating freely available DEM datasets that may be suitable for this purpose. In parallel, the student will become familiar with GNSS fundamentals and RAIM prediction algorithms.

The second phase will involve the implementation of a terrain-based elevation masking angle computation, including an in-depth analysis of the trade-off between estimation accuracy and computational cost. The implemented algorithm will need to be optimized to avoid an unnecessary high number of computational iterations that do not provide any additional benefit in terms of estimation precision for the intended purpose.

Finally, the last phase will consist of integrating the terrain-based elevation masking algorithm into an existing RAIM prediction algorithm, which is currently limited to the use of fixed mask angle values.
 
Overview (resum en anglès):
This project designs, develops and integrates an optimized algorithm to generate an elevation angle mask dependent on the terrain's orography through Digital Elevation Models (DEM), increasing the accuracy of the RAIM (Receiver Autonomous Integrity Monitoring) prediction system previously developed by PildoLabs. In current air navigation, specifically under PBN procedures, GNSS integrity is of vital importance. Standard algorithms use fixed-value masks (typically 5º) that ignore terrestrial obstacles, increasing the risk of controlled flight into terrain (CFIT) accidents due to overestimating satellite availability at low altitudes. This thesis addresses this problem by analyzing the actual horizon in all directions from the receiver antenna onboard the aircraft.
The methodology was divided into three phases. First, the global public model Copernicus DEM GLO-30 (resolution of 30 meters) was selected, after carrying out an exhaustive study of the different existing Digital Elevation Models. Second, a modular software architecture was designed in Java, implementing a geodesic ray-casting algorithm. To reduce the high computational cost of processing large volumes of elevation data, various optimizations were introduced: a cache to store already loaded maps, filters based on maximum theoretical distances, oceanic optimization, manual bilinear interpolation, and an Early Stopping criterion. Third, the module was integrated into the RAIM engine and validated in three orographically distinct scenarios: Andorra, Ciudad Real Airport (LERL) and the UPC Campus Nord.
The system was validated using three different tools. QGIS allowed a geometric validation to ensure that the elevation profiles generated by the algorithm match the real DEM data. HeyWhatsThat confirmed that the masks obtained with the program faithfully represent the actual horizon. GNSS Mission Planning verified the correct integration of the module into the RAIM prediction engine by checking hourly satellite visibility.
Regarding performance, parallel map preloading accelerated loading by a factor of 3.51 for local repositories and 6.03 for files obtained in real-time via AWS. The Early Stopping criterion significantly reduced computational time, especially in mountainous areas like Andorra, where processing almost stops immediately. Additionally, analyzing multiple waypoints on the same route showed highly optimized performance: the time per point decreases considerably as their number increases, needing only 200 ms per point when processing 20 waypoints compared to the 1500 ms required with a single point.
In mountainous scenarios such as Andorra, ignoring terrain in RAIM prediction is not a viable simplification: satellites that would actually be blocked are validated, increasing the risk of unexpected integrity loss during the flight. The developed module directly corrects this problem, matching the pre-navigation prediction with the situations the receiver will encounter en route.

© CBLTIC Campus del Baix Llobregat - UPC