Quality Analysis of GNSS Observations and Real-Time Precise Point Positioning for Wearable Smart Devices: A Case Study of Honor Watch GS3

With the growing demand for high-precision location-based services (LBSs) with the Internet of Things (IoT) smart terminals, realizing high-accuracy positioning with low-cost devices has become a key focus in the global navigation satellite system (GNSS) research field. Wearable smart devices, as widely used low-cost IoT terminals, hold significant application potential in LBSs. This study employs the Honor Watch GS3 as a platform to systematically analyze the quality of GNSS data and investigate its potential for achieving real-time high-precision positioning. First, the quality of the raw GNSS observations from the smartwatch is analyzed and compared with that obtained from Huawei Mate40 and Xiaomi Mi8 smartphones. The study shows that an average of 30.5 satellites was observed per epoch by the smartwatch, with an average carrier-to-noise density ratio (C/N0) of 33.37 dB $\cdot $ Hz. The root mean square (rms) of pseudorange observation noise for the smartwatch is approximately 3.23 m, and the carrier phase noise is approximately 0.07 m. Overall, the GNSS observation quality of the wearable smart device exhibits characteristics similar to those of smartphones, but with uncommon phenomena such as receiver clock jumps in pseudorange observations and half-cycle ambiguities in Galileo satellite observations. Then, a smartwatch-specific precise point positioning (PPP) model is developed with a motion constraint applied to optimize the positioning performance, and PPP is achieved using real-time service (RTS) and PPP-B2b. Finally, kinematic experiments in a campus environment were conducted to validate the proposed method, demonstrating significant accuracy improvements compared to the embedded solutions of the experimental smartwatch. The PPP results using RTS showed that the horizontal accuracy has improved by approximately 40% in items of mean value (MEAN), rms, and the 95th percentile error (P95). For the 3-D accuracy, improvements of 45%, 42%, and 35% were observed for MEAN, rms, and P95, respectively. The results using PPP-B2b also showed improvements in horizontal accuracy, with gains of about 30% for MEAN, rms, and P95. In terms of 3-D accuracy, the improvements were 37%, 34%, and 26% for MEAN, rms, and P95, respectively. These results confirm the effectiveness of the proposed method in enhancing smartwatch real-time positioning accuracy.