High Accuracy GNSS Augmentation With PointPerfect – A New Approach for Enabling Mass Market cm-Level Positioning
Rodrigo Leandro, u-blox Michael Albright, u-blox Landon Urquhart, u-blox Kendall Ferguson, u-blox Hunaiz Ali, u-blox Andre Ebeling, u-blox, Dragos Catalin, u-blox
Date/Time: Wednesday, Sep. 18, 4:23 p.m.
Over the last several years the GNSS community has been progressively relying on high accuracy corrections made available in broadcast form, i.e., in one-way communication systems. There are several reasons for the popularity of the broadcast mode for GNSS, including for instance the compatibility with satellite communication, where two-way communication is still either not possible or economically prohibitive. Although we expect this scenario to change overtime with the advent of LEO-based internet, this limitation is still a reality for GNSS correction broadcast over communication satellites.
The GNSS correction data transmitted in broadcast mode has an important core requirement for usability – being ubiquitous. This means that the data must be usable by any device that is receiving it, regardless of the user location. Of course, this requirement is under the scope of the coverage area the corrections are intended and designed for. Nevertheless, corrections must be usable according to accuracy specification for any location under the service coverage region definition. The extent of the coverage of any given service can vary widely, from local to global scale depending on the purpose of the service. In case of satellite transmission, the coverage area for the transmitted data is typically restricted to a continental scale, because it does not to extend beyond the range of the communication satellite itself.
The ubiquity requirement is not a new development in GNSS and has been explored for a few decades now. The way this is resolved is by separating each aspect of the satellite signal characteristics, in a way that the user can assemble the proper components using the proper assumptions for their location. There are two main categories of the signal characteristics: corrections and models. The corrections are effectively the information that is carried in the correction data payload. This includes all aspects that need to be explicitly communicated to the user and can include satellite orbit parameters (e.g. satellite position and attitude information), satellite hardware behavior parameters (e.g. satellite clock and hardware delay behavior). The models are the physical and mathematical models that both the service and the user rely on to fully represent the satellite signals in a mathematical model. They typically include known or well-approximated effects that can be calculated without having to rely on real time data information, and can include everything from geodynamic effects (e.g. planetary tides, and earth rotation parameters) to satellite attitude-driven effects on carrier waves (phase wind-up effects). Also importantly, the consistency of models between service and user has a direct impact on the signal modelling accuracy, therefore somewhat representing a contract between service and user on what the physical and mathematical assumptions are in place for the correction data in question.
Because of the compatibility with satellite data transmission, broadcast correction models have been used for decades now, and have grown in popularity with advances in science and technology that allowed us to create positioning systems with centimeter level accuracies on continental and even global scale. As a result of this, many of the high accuracy commercial and government services rely on broadcast data models and formats, some of them public and some proprietary. The importance of this tendency is that the service and rover systems have also been progressing towards using broadcast models as the default way of communicating and using high accuracy data. Because these are reasonably complex systems, it is convenient to consolidate around a given model of communication, as different model variants might require extensive adaptation and tuning of the engines within the system framework.
Let’s consider a system using the broadcast model as our basis of discussions here and explore another aspect – communication options and redundancy. We discussed many aspects of satellite communication; however, we must consider that the augmentation data is often sent over land-based communication (e.g. land-based internet) as well. In general terms this is very straightforward since we can simply pipe the same data that was uploaded to the satellite through the internet. However, while simple, this solution generates a massive amount of wasted bandwidth. This is because the data payload carries enough information to serve any user in the satellite coverage, and therefore any given user will employ only a fraction of the data payload for their purpose. In traditional applications and services this has not been a problem at all, because the bandwidth cost was a minimal fraction of the overall cost of hardware and correction service. When operating in the IoT domain some of those premises are extremely different. The receiver hardware cost is much lower (even though it is still capable of delivering high accuracies), and the service cost is also down to comparable levels of data costs. The service economics change because of volume of users, which allow massive improvement in shared cost for that user base. Therefore, a new problem was established for the IoT-grade high accuracy GNSS operators, the cost optimization for broadcast data models over internet.
Solving the bandwidth optimization problem involves a range of low hanging fruit solutions and some other that turned out more complex as we move towards further optimization, in combination with ensuring that certain aspects of the broadcast model are kept. One of the key premises for broadcast models is that the service does not know where the user is. This premise is particularly important in situations where user data privacy must be maximized. Consequently, the bandwidth efficiency involves two main classes of technical solutions, the data reduction strategy; and the user privacy strategy, combined. Those solutions are explored in this paper.
While bandwidth is important, accuracy performance is one of the most critical aspects of positioning. To understand positioning accuracy performance, we need understand the data modeling challenges of a GNSS satellite to receiver measurement. To achieve positioning accuracies at the cm level some of the effects of the signal must be known, eliminated, or calculated within or close to that level. Depending on the nature of the effect, different approaches are more appropriate. For instance, the position of the satellites in an earth-centered coordinate system are difficult (if not impossible) to be calculated by a standalone receiver. Therefore, this aspect is better handled by either elimination with using data from a nearby receiver at known location, or with explicit calculations performed by a network of receivers. In addition to that, because the satellite motion is driven by well-known physics, it is typically easy to model over longer periods of time. Satellite clock behavior, as opposed to orbits, is more difficult to model over time, and its predictability depends on the quality of the clock standard being employed by each spacecraft. The temporal behavior of satellite clocks is one of the main drivers on data update rates for correction systems since those need to keep sending data at a high enough rate to represent the current state of the satellite clock within a certain accuracy. The satellite-based effects such as the orbits, clocks, and equipment measurement biases are all geometry independent and therefore have a common impact on receiver measurements, independent from where these receivers are on the planet. This is because these effects are the basis of corrections systems that provide data applicable at a global scale, often referred as PPP. Also, their calculation has limited dependency on the density of the tracking networks because accurate satellite effects can be calculated whenever the satellite is observed by a few receivers observing the satellite from different directions (different enough to provide decorrelation between variables of a linear system estimation model) at the same time.
The atmosphere effects differ from the satellite effects in several ways. This is partially because the atmosphere observed by any given receiver is just a fraction of the general atmosphere represented in a correction model payload delivering high accuracy performance. The atmosphere data size often represents 90% or so of the data payload size for continent-wide corrections. This amount of data is required because the atmosphere has irregular behavior both in terms of space and time. Satellite behavior description has no correlation with space, since it can be used ubiquitously, and with the exception of satellite clock stability is also fairly well behaved over time. While the atmosphere is still better behaved than satellite clock errors in the time domain, it still requires a larger amount of data to properly represent all the ground of the broadcast representation. While difficult to represent, the atmosphere, specially the ionosphere, is a fundamental part of enabling a rover receiver to calculate high accuracy positions, specially in difficult tracking environments (this relationship is covered later in the text). Because of the spatial variability of the atmosphere, its representation is heavily correlated with the density of the tracking network used to calculate its impact on GNSS signals. The ionospheric effects are typically resolved on the user end with a combination of tracking GNSS signals on multiple frequencies (ionosphere is a dispersive medium and therefore its impact relates to signal frequency) and a reasonably well-known initial state typically acquired from an ionospheric representation model provided by the correction data system in use.
In this paper we are presenting a new correction service system that has been designed and built from ground up to best serve users in the IoT domain. This service is the second generation of u-blox’ PointPerfect. In the process of architecting the new service, each aspect of correction calculation, usage, and transmission was considered to optimize the several aspects of performance for today’s as well as future IoT applications and connectivity technology. In this process we abandoned some of the hard requirements of version one, such as strict homogeneous performance, and also established new strong goals such as scalability and cloud computational efficiency. While most systems we (GNSS professionals, and industry in general) have designed over the last decades have been created aiming at satellite broadcast distribution systems, the new PointPerfect was architected to serve the future of IoT. Nevertheless, the new service still attends every requirement of its predecessor. While introducing several improvements and new ways of seeing the future of GNSS high precision usage in the world. Among other things, the new system enables cm-level accuracies with initialization times down to 1-2s depending on the use case. It is also built on a more efficient software framework that allows processing times of under 500 ms between network data observation collection and transmission to rover. Compared to its predecessor, the overall computational load is improved by more than 90% which makes it quite more energy efficient, while enabling unprecedented scalability for supporting unusually high-density networks whenever applicable. While the core accuracy performance is better than before, the rover robustness in difficult tracking environment is also improved with the new system higher resolution ionospheric data. Therefore, the effective performance in real life situations is improved further than the improvement in nominal accuracy. The relationship between nominal accuracy, tracking environment, and effective accuracy are explored in more detail in the full paper.
In summary, the new system delivers unprecedented user experience, which has been made possible by the innovative design, and more importantly, the different way to see user’s needs and connectivity environment now and in the future. This work explains all the fundamentals behind the design of this new system and explores the architecture that has allowed the achievement of the new performance metrics. Results obtained both in controlled environment and in the field are also shown and analyzed, with comparisons to previous and existing alternatives when applicable.
For Attendees Call for Abstracts Registration Hotel Travel and Visas Exhibits Submit Kepler Nomination For Authors and Chairs Abstract Management Student Paper Awards Editorial Review Policies Publication Ethics Policies For Exhibitors Exhibitor Resource Center Marketing Resources Other Years Future Meetings Past Meetings