Hybrid Positioning Framework with Innovation Based Judgment in Complex Indoor Environments

In indoor environments, complex structures not only introduce Line of Sight (LOS) paths, but also give rise to Non-Line of Sight (NLOS) paths. This scenario presents immense challenges for positioning technologies. Currently, most existing indoor positioning solutions primarily target either LOS or NLOS environments and employ a singular positioning algorithm, which leads to suboptimal robustness in positioning. In order to tackle this issue, this study proposes a novel indoor hybrid positioning scheme. This model unifies multiple positioning algorithms, allowing each to leverage its advantages in either LOS or NLOS environments. The model, based on the innovation residual judgement result of the Maximum Correlation Entropy Kalman Filter, integrates the output positions from multiple positioning models, thereby adaptively selecting the optimal positioning method. In the event of Line of Sight (LOS) conditions, the model opts for the positioning results obtained from an enhanced CHAN-Kalman position estimation method based on Time Difference of Arrival (TDOA). Conversely, under Non-Line of Sight (NLOS) conditions, we propose a Convolutional Neural Network (CNN) fingerprint positioning method based on 5G channel impulse response for position estimation. Upon verification on an open-source dataset, the method proposed in this study can achieve single-point positioning in both LOS and NLOS mixed environments. Compared to single positioning methods, it possesses a higher degree of positioning accuracy.