Combined Algorithm for Satellite Selection for Open-sky and Constrained Environments

Abstract

The objective of this research is to develop an algorithm for satellite selection that considers both satellite geometry and measurement quality under various positioning environments with a desired computational time and equivalently good positioning results. Multi-constellation global navigation satellite systems (GNSSs) are often integrated to improve positioning performance. Compared to stand-alone systems, multi-constellation GNSSs provide more opportunities for users to receive sufficient signals. An increase in the number of visible satellites leads to a higher chance of generating a position solution and improves satellite geometry. Therefore, the accuracy, continuity, and availability of the positioning system can be enhanced. However, the total global satellite number will soon increase to almost 150. In the Asia Pacific region, signals from multiple GNSSs (GPS, GLONASS, and BeiDou) are commonly received, with users often simultaneously receiving signals from 50 satellites. Although increasing the number of visible satellites improves positioning performance, it also increases computational load. For low-cost commercial receivers, this is a serious issue since computation resources are restricted and the number of tracking channels is limited. Satellite selection algorithms can be utilized to address this issue. Satellites are selected and the position solution is computed using a subset of the received satellite signals. In constrained environments, such as urban and mountainous areas, it is a challenge for GNSSs to provide positioning service for users. One major difficulty is the reception of multipath and non-line-of-sight (NLOS) signals. Such contaminated signals caused by reflections off of obstacles such as buildings and terrestrial objects may generate positioning errors of up to 100 m without being detected by a conventional portable device. In urban areas, users cannot identify the correct route since this magnitude of error exceeds the width of the road. Research into utilizing receiver autonomous integrity monitoring (RAIM) fault detection and exclusion (FDE) to mitigate the multipath and NLOS signal effects have increased. RAIM FDE is designed to identify outliers within the signal set to detect and exclude both multipath and NLOS signals. Conventional RAIM FDE algorithms are based on a single-fault assumption. However, in constrained environments, receivers are likely to receive multiple contaminated satellite signals. Thus, RAIM FDE with a multiple-fault assumption is more suitable for users in constrained environments. Additionally, an extended Kalman filter (EKF) can be used to smooth positioning results. Since map information is stored in modern portable navigation devices, one can regard urban roads as the assembly of piecewise continuous lines and use the map information to build road model equations. These equations can then be applied in the EKF prediction process to constrain user positions along the desired path. In addition to immunity to outlier error, the extra equations and parameters derived from the road model are beneficial to users in scenarios with insufficient satellites. This can reduce the minimum numbers of received satellites and minimum signals numbers required to operate RAIM FDE. Therefore, the availability of positioning results as well as RAIM FDE is increased by the road-model-constrained EKF solution. In this work, two satellite selection algorithms that consider satellite geometry are used, namely quasi-optimal satellite selection (QOSS) and tri-angle optimal (TAO) satellite selection algorithms. Receiver autonomous integrity monitoring (RAIM) fault detection and exclusion (FDE), with a multiple-fault assumption, based on greedy search is used for checking measurement quality. For users in open-sky areas, two rounds of the geometry satellite selection and the measurement quality check are conducted and only six satellites are needed for calculating the position solution. Since three constellations of a global navigation satellite system are utilized in this study, the user position in three directions and three user clock biases should be calculated, a minimum of six satellites are required for integrated positioning for a three multi-constellation GNSS. The satellite selection algorithm performance was evaluated using a conventional receiver with static and dynamic experiments. The results show that the satellite selection algorithms used in this research can select a measurement set with good horizontal dilution of precision (HDOP) value, horizontal positioning error (HPE), and relatively clean signals within a reasonable time. For users in constrained environments, one geometry selection is performed followed by RAIM FDE. An extended Kalman filter with a road-model-constrained solution is adopted in extreme positioning scenarios. The method effectively generates accurate positioning results for users. Finally, the algorithm is evaluated using a real dataset collected with a conventional receiver without additional sensors, and is compared with other positioning algorithms to test its applicability. The results show that this positioning algorithm can efficiently provide adequate accuracy for users.