Content area
Ultra-wideband (UWB) is a promising indoor positioning technology with high accuracy and low energy consumption. The current research uses UWB, assisted by wheeled odometers, to develop a stable centimetre-level indoor positioning system (IPS) to address smart industry needs.
The research initially investigates the limitations of UWB-only IPS and identifies non-line-of-sight (NLOS) errors caused by signal occlusion as the primary factor affecting positioning accuracy. The main goal of this research phase is to identify and reduce NLOS errors accurately. We propose a sliding window algorithm that can identify about 90% of NLOS errors by this very simple algorithm in a very harsh experimental environment. This algorithm can accurately identify NLOS by analyzing only the variation of ranging values within a sampling window without analysing parameters such as signal strength or channel impulse response (CIR). The algorithm is simple and can be applied to most UWB devices. Additionally, we present a method for modelling NLOS errors induced by walls based on wall thickness, material properties, and UWB signal incidence angle to mitigate such errors effectively within room structures. Experimental verification demonstrates that accurate identification and modelling of NLOS errors can significantly improve UWB-only IPS accuracy under NLOS conditions close to line-of-sight (LOS) levels. Furthermore, our research explores strategies to reduce the cost of UWB hardware in large-area scenario applications by proposing a method to convert fixed anchors into moving ones.
In flexible, variable, and complex indoor environments, achieving accurate positioning with UWB alone is difficult. This research combines low-cost wheeled odometers and UWB to form a multi-sensor fusion localisation system. We proposed a loosely coupled method that fuses UWB and odometry. The first step is to identify the NLOS using the sliding window method and then optimise the positioning of the UWB system by deleting ranges that contain NLOS errors. Finally, the absolute position of the optimised UWB is selected to correct the cumulative error in the direction and displacement of the odometer for precise positioning. The fused system can achieve a positioning accuracy of less than 10 cm RMSE in complex environments. Loose coupling requires subsystems to realise localisation independently, which may not be possible for UWB systems in harsh NLOS environments, affecting the fused system's accuracy. The final phase of the research proposes a tight coupling method based on the dynamic Unscented Kalman filter (UKF), which assists the UWB in identifying and mitigating the NLOS errors and calculates the Horizontal dilution of precision (HDOP) value through the position information provided by the odometer. The parameters and inputs of the UKF are dynamically adjusted according to the NLOS state, HDOP value, and motion state to achieve stable and accurate positioning with RMSE less than 10 cm in strong NLOS environments.