A K-NN-Driven Multilateration Approach for Improved Aircraft Positioning

Abstract

For safe and efficient air traffic management, the Air Traffic Control (ATC) should know the precise location of aircraft. Aircraft usually report their positions to ATC using an advanced location-based service known as Automatic Dependent Surveillance–Broadcast (ADS-B). The location of aircraft without position-reporting capabilities is determined using complementary localization methods. A key challenge with traditional positioning techniques, such as multilateration using Time Difference of Arrival (TDOA), is that they involve solving non-linear equations, which require a precise initial position estimate. In this paper, we propose a novel method for aircraft localization that integrates a traditional positioning technique (multilateration) with data-driven learning using the K-Nearest Neighbors (K-NN) algorithm. The K-NN regression model provides a more realistic initial guess of the aircraft’s position. The results were validated against the actual aircraft positions provided by the OpenSky Network, and the proposed technique demonstrated a 2D root-mean-square error of 39.4 m. This work has significant potential for real-world applications in air traffic management, contributing to safer and more precise aircraft positioning.