2:45 PM - 3:00 PM
[MTT37-05] Introduction to Quality Management Methods for Dense GNSS Observation Networks
Keywords:GNSS, LOS, NLOS, machine learning
Due to the impact of labor shortage caused by population decline in Japan, automation has been increasingly adopted in various fields such as agriculture and construction. One of the necessary technologies for achieving automation is high-precision self-positioning, and high-precision positioning services using GNSS have gained attention. SoftBank Corp. has deployed the high-precision positioning service "ichimill" using original reference sites installed in over 3,300 locations nationwide, and delivers correction information required for RTK positioning.
To ensure the quality of the service, it is necessary to regularly and comprehensively verify the over 3,300 original reference sites. We place particular importance on the accuracy degradation of observation values caused by changes in the surrounding environment of the reference site.
One of the factors contributing to the degradation of accuracy is the line of sight (LOS). When there is no degradation in the observation accuracy, it means that there is a clear line of sight (LOS) between the satellite and the antenna. However, when there are obstacles such as buildings and vegetation between the satellite and the antenna, the line of sight can be obstructed, resulting in a non-line of sight (NLOS). In the case of NLOS, the received C/N ratio of observation values may decrease or become unreceivable. Previously, the LOS/NLOS determination was made by image recognition from photographs taken above the reference sites. However, accurately determining the azimuth and elevation angles from the captured images proved difficult, resulting in inaccurate determinations. Additionally, conducting accurate determinations required temporarily shutting down the operation of the reference sites to capture photographs above the antennas, which posed problems in terms of cost and maintaining the operational rate of the reference sites. Therefore, we employ machine learning to determine LOS/NLOS based on the C/N ratio obtained from observation data at the reference site. and the azimuth and elevation angles of the satellites, effectively resolving these problems. This determination is periodically conducted for all reference sites, ensuring the monitoring of their quality.
Furthermore, we are also conducting research and development on alternative monitoring methods for factors other than the line of sight, aiming to provide high-quality correction data distribution services.
To ensure the quality of the service, it is necessary to regularly and comprehensively verify the over 3,300 original reference sites. We place particular importance on the accuracy degradation of observation values caused by changes in the surrounding environment of the reference site.
One of the factors contributing to the degradation of accuracy is the line of sight (LOS). When there is no degradation in the observation accuracy, it means that there is a clear line of sight (LOS) between the satellite and the antenna. However, when there are obstacles such as buildings and vegetation between the satellite and the antenna, the line of sight can be obstructed, resulting in a non-line of sight (NLOS). In the case of NLOS, the received C/N ratio of observation values may decrease or become unreceivable. Previously, the LOS/NLOS determination was made by image recognition from photographs taken above the reference sites. However, accurately determining the azimuth and elevation angles from the captured images proved difficult, resulting in inaccurate determinations. Additionally, conducting accurate determinations required temporarily shutting down the operation of the reference sites to capture photographs above the antennas, which posed problems in terms of cost and maintaining the operational rate of the reference sites. Therefore, we employ machine learning to determine LOS/NLOS based on the C/N ratio obtained from observation data at the reference site. and the azimuth and elevation angles of the satellites, effectively resolving these problems. This determination is periodically conducted for all reference sites, ensuring the monitoring of their quality.
Furthermore, we are also conducting research and development on alternative monitoring methods for factors other than the line of sight, aiming to provide high-quality correction data distribution services.