– Accurate and traceable drone positioning is crucial for autonomous aerial navigation, especially in GNSS-denied environments. Traditional approaches using Ultra-Wideband (UWB) sensors or Kalman Filters (KF) struggle with multipath interference, non-line-of-sight, and environmental uncertainties, and are often limited to Linux-based Application Programming Interfaces (APIs). This work presents an innovative framework based on: a novel modular Windows-based API for real-time drone positioning, the integration of Kalman filter for optimal multi-sensor data fusion and trajectory smoothing, AI-driven residual learning to correct systematic estimation errors, and metrology-compliant uncertainty modeling. The system enables real-time swarm deployment and pose-aware feedback using an auxiliary vision based positiong system (OptiTrack) and UWB data. A feedforward neural network compensates for residual errors in Kalman-filtered trajectories, while Monte Carlo simulations establish traceable 95% confidence intervals. Experimental tests show that the proposed framework reduces RMSE by over 40% across axes, with strong regression accuracy greater than 94%.
A Modular Windows-Based Intelligent API for Traceable Drone Positioning Using UWB-OptiTrack Fusion and AI-Based Residual Learning
Haq I. U.
;D'Alfonso L.;Fedele G.;Lamonaca F.
2025-01-01
Abstract
– Accurate and traceable drone positioning is crucial for autonomous aerial navigation, especially in GNSS-denied environments. Traditional approaches using Ultra-Wideband (UWB) sensors or Kalman Filters (KF) struggle with multipath interference, non-line-of-sight, and environmental uncertainties, and are often limited to Linux-based Application Programming Interfaces (APIs). This work presents an innovative framework based on: a novel modular Windows-based API for real-time drone positioning, the integration of Kalman filter for optimal multi-sensor data fusion and trajectory smoothing, AI-driven residual learning to correct systematic estimation errors, and metrology-compliant uncertainty modeling. The system enables real-time swarm deployment and pose-aware feedback using an auxiliary vision based positiong system (OptiTrack) and UWB data. A feedforward neural network compensates for residual errors in Kalman-filtered trajectories, while Monte Carlo simulations establish traceable 95% confidence intervals. Experimental tests show that the proposed framework reduces RMSE by over 40% across axes, with strong regression accuracy greater than 94%.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.


