A Novel Prediction Technique for Intermittent UWB Positioning System Using Hybrid LSTM-Trilateration

Accurate positioning has become more critical for elderly care applications, particularly in indoor environments where Global Navigation Satellite Systems (GNSS)-based localization is usually unavailable. Ultra-wideband (UWB) radio frequency technology, with centimeter-level accuracy, offers a robust solution without requiring physical connections. However, wireless applications face challenges from intermittent signal loss due to factors like electromagnetic interference or multipath propagation, posing risks to vehicle operations. With increasing computational power, LSTM networks stand out as effective tools for modeling complex temporal relationships in UWB positioning data, leveraging both current conditions and historical data for accurate predictions and reliable navigation in dynamic environments. In this paper, we propose a hybrid LSTM-Trilateration method for direct and indirect predictions. The final results show that both methods have similar performance, with the best being 1.4 cm error. Moreover, compared to other works, the proposed method achieves the smallest error.