UWB-based Machine Learning Optimized 3D Positioning Algorithm

Conventional UWB-based positioning algorithms require multiple base stations to be installed in different corners of the room, enabling accurate 3D positioning. In this paper, we design a miniaturized indoor positioning device that has nearly no height difference between the base stations. By considering several machine learning optimization algorithms, we design a hybrid method which combines the Levenberg-Marquardt algorithm and TOA positioning algorithm that can precisely locate the height of a target in this situation. Hybrid method is more accurate than the traditional TOA and TDOA method, with no significant performance degradation in computing speed.