Evaluation of Dynamic Measurements from RTK-GNSS Operating in Sub-urban Environment

• Published in: 2021 IEEE 12th Control and System Graduate Research Colloquium, ICSGRC 2021 - Proceedings
• Main Author: Nahar R.S.; Ng K.M.; Johari J.; Kamal M.M.; Razak N.A.
• Format: Conference paper
• Language: English
• Published: Institute of Electrical and Electronics Engineers Inc. 2021
• Online Access: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85114610589&doi=10.1109%2fICSGRC53186.2021.9515275&partnerID=40&md5=1393070a1b0e31f44be57ff4dbfb259f

Previous works that evaluated the performance of the EMLID Reach RS RTK-GNSS based on its accuracy and availability may not be sufficient to reveal further noise characteristics in the residual. Hence, it is noteworthy to further investigate correlated noise characteristics and the probabilistic distribution in the error residuals to assist in appropriate filtering and statistical description of the noise. In view of this, the authors aim to evaluate measurements from the EMLID Reach RS RTK-GNSS in terms of accuracy, availability, correlated noise characteristics and probabilistic distribution of the errors when operating dynamically in a sub-urban environment. Data logging was performed with the rover moving in two routes located in a residential area in Shah Alam, Malaysia with baseline of less than 200 m and above 200 m from the base station respectively. Seven sets of data were collected for both routes. The data collected was analyzed in comparison with ground truth to investigate accuracy and availability. Subsequently, the errors were evaluated using autocorrelation and partial correlation tests to reveal characteristics of correlated noise. A distribution fitter was used to identify the best type of distribution that can fit the error residuals. In terms of accuracy, MAE and RMSE are better in dynamic measurements with baseline less than 200 m compared to above 200 m. The autocorrelation and partial correlation plots for all data sets indicated the presence of correlated noise in the error residuals, which could be described using the non-parametric distribution. Hence, non-parametric probability density could be used to estimate the mean and the bounds in the error. On the other hand, appropriate filtering is needed to attenuate correlated noise. © 2021 IEEE.