Performance evaluation on extended neural network localization algorithm on 5 g new radio technology.

With the rapid growth of fifth-generation (5G) networks, there is an increasing demand for high-precision localisation, achieving which is a major challenge in real-time applications in dynamic and noisy environments. Signal noise and incomplete data, including time difference of arrival (TDoA), angle of arrival (AoA), and frequency of arrival (FoA), often limit traditional methods from achieving improved localization. This research proposes an advanced hybrid localisation method combining Extended Kalman Filter (EKF) and Extended Neural Network (ENN) with HackRF-based software-defined radios (SDRs) to improve the real-time localization in 5G environments. The method achieves a better localization accuracy by using EKF for noise reduction and ENN for localization data fusion and combining FoA, AoA, and TDoA measurements. Experiments using real-time 5G signal data show that the proposed EKF-ENN fusion outperforms the existing methods. It obtains an AoA mean of 0.08 radians (SD = 0.014 rad), a TDoA mean of 0.020 s (SD = 0.003 s), and a FoA mean of 0.49 Hz (SD = 0.09 Hz). Its Mean Squared Error (MSE) of 1.06e and Signal-to-Noise Ratio (SNR) of 11.7 dB show that it attains better performance than existing ones. Its increased localisation accuracy and signal processing efficiency qualifies it for real-time usage in next-generation wireless networks.