Kalman Filter Optimization of Anchor Location in a GPS-denied Relative Network
Authors
Ghanim, Muad Naim ; Larsen, Benjamin Speyer
Term
4. term
Education
Publication year
2026
Submitted on
2026-06-11
Abstract
This thesis investigates how an existing indoor localization framework, the IMU-Controlled Ultra-Wideband Mesh (I.U.M.), can be extended to link relative positions to geographically meaningful coordinates under GPS-limited conditions. The I.U.M. system allows mobile nodes to determine their mutual positions using peer-to-peer UWB ranging and inertial measurements without fixed anchors, but it only provides a local relative coordinate frame. To address this, the work introduces a limited-GPS anchor approach in which a single node is initialized with a GPS-derived geographic coordinate that is propagated through the UWB mesh so that other nodes obtain approximate global positions. Because the GPS-based anchor may be noisy, an adaptive Kalman filter is implemented to estimate and smooth a shared position bias before geographic information is distributed across the network. The filter continuously adjusts its process and measurement noise covariances using innovation-based feedback to better cope with time-varying measurement uncertainty. The extended simulator is evaluated purely through simulations, including representative runs and Monte Carlo trials. The results show that the adaptive Kalman filter reduces short-term fluctuations and improves temporal stability of the position estimates, but improvements in absolute positioning accuracy are inconsistent and the RMSE is not reliably reduced across all runs. The thesis concludes that, in its current form, the method is more effective as a smoothing mechanism than as a generally robust accuracy-enhancing localization technique, and it highlights the need for richer dynamic inputs, real-world experiments, and improved adaptive tuning in future work.
Denne afhandling undersøger, hvordan et eksisterende indendørs lokaliseringssystem, IMU-Controlled Ultra-Wideband Mesh (I.U.M.), kan udvides til at koble relative positioner til geografisk meningsfulde koordinater i GPS-begrænsede miljøer. I.U.M. gør det muligt for mobile noder at bestemme deres indbyrdes positioner via UWB-afstandsmålinger og inertielle sensorer uden faste ankre, men leverer kun et lokalt relativt koordinatsystem. Arbejdet introducerer en begrænset-GPS-tilgang, hvor én enkelt node initialiseres som anker med en GPS-afledt geografisk koordinat, som derefter forplantes gennem UWB-meshet, så de øvrige noder kan opnå omtrentlige globale positioner. Da GPS-ankeret kan være støjfyldt, implementeres en adaptiv Kalman-filterløsning, der estimerer og udglatter en fælles positionsbias, før de geografiske oplysninger deles i netværket. Filteret justerer løbende sine støjkovarianser baseret på innovationsfeedback for bedre at håndtere tidsvarierende måleusikkerhed. Den udvidede simulator evalueres udelukkende ved hjælp af simuleringer, herunder repræsentative kørsler og Monte Carlo-forsøg. Resultaterne viser, at det adaptive Kalman-filter reducerer kortsigtede udsving og forbedrer den tidsmæssige stabilitet i positionsestimaterne, men at forbedringer i absolut positioneringsnøjagtighed er inkonsistente, og at RMSE ikke pålideligt mindskes i alle forsøg. Afhandlingen konkluderer derfor, at den nuværende løsning primært fungerer som et glatningsredskab frem for en generelt robust metode til at øge absolut nøjagtighed, og peger på behovet for rigere dynamiske input, praktiske eksperimenter og bedre adaptiv tuning i fremtidigt arbejde.
[This abstract has been generated with the help of AI directly from the project full text]