The design of a GNSS-based integrity concept for automotive applications presents many challenges that are related to the complicated propagation channel encountered in a typical road environment. The presence of obstacles and/or reflectors on or beside the road such as large trucks, buildings, trees, walls, overpasses or tunnels can result in a large variation in the number of received signals or in the quality of the GNSS measurements and have a negative impact on the Gaussianity of the error distribution and the independence of these errors both in time and across measurements. u-blox has been working, among other solutions, on a new integrity concept referred to as Single Epoch Position Bound (SEPB) [1, 2]. It is built on: • a Bayesian estimation framework • the use of multi-frequency multi-GNSS pseudorange and carrier phase measurements • a snapshot position computation to avoid modeling time correlation • highly dynamic non-Gaussian error modeling SEPB has been shown to provide particularly tight bounds on the position. Non-Gaussian GNSS measurement error models have some major advantages in the context of high-integrity automotive positioning. For the same integrity risk, non-Gaussian error models can yield tighter position bounds than traditional Gaussian overbounding techniques, and in addition the mathematical analysis is simplified. The only major disadvantage is that the position bound evaluation can be computationally demanding, particularly if there are many unknown nuisance parameters which need to be included in the state. This is because the Bayesian framework provides the posterior distribution of the position as a function of a potentially large number of states, and evaluating bounds thus results in a computationally intensive numerical integration. To some extent, it is possible to mitigate this by computing the SEPB bound at low frequency (e.g. once every few seconds) and propagating the bound in between, but even with this optimization the computational requirements are still high. One major area for concern when using code and phase measurements in SEPB is the ionosphere. Using ionosphere-free measurements is possible but undesirable because SEPB then cannot take advantage of phase measurements. Indeed, SEPB does not fix integer phase ambiguities, but rather it integrates over the integer ambiguities when forming bounds, and hence it benefits greatly from phase measurements. There is thus a need to model the ionospheric delay. Due to spatial variations, it is difficult to safely model the ionosphere with anything other than a per-satellite parameter, which naturally leads to a large number of unknown states in the system when multiple GNSS are used. Previous published work on SEPB has side-stepped this problem by relying on short baseline RTK corrections, which made ionospheric errors negligible. However, for the non-Gaussian approach to be useful in real-world products it is essential that non-negligible ionospheric effects can be accommodated. In this paper we describe and evaluate advances in SEPB which includes per-satellite ionosphere states. We find that adding per-satellite ionosphere states does result in some increase in computational load, but that with careful design of the numerical integration scheme this load is still tractable for real-time applications. The use of bound propagation between SEPB solutions is shown to lead to an increase of the bound that remains acceptable. The bounding performance is finally evaluated based on a significant amount of real road data together with the use of a PPP correction service, to show that useful position bounds can be obtained without any atmospheric corrections. [1] Bryant, R., Julien, O., Hide, C., Moridi, S., Sheret, I., "Novel Snapshot Integrity Algorithm for Automotive Applications: Test Results Based on Real Data," 2020 IEEE/ION Position, Location and Navigation Symposium (PLANS), Portland, Oregon, April 2020, pp. 670-681. [2] Bryant, Rod, Julien, Olivier, Hide, Chris, Skorupa, M., Sheret, Ian, "Road Vehicle Integrity Bound Propagation Using GNSS/IMU/Odometer," Proceedings of the 33rd International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2020), September 2020, pp. 585-611.

The design of a GNSS-based high integrity navigation system for automotive applications presents many challenges that are related to the complicated propagation channel encountered in a typical road environment. The presence of obstacles and/or reflectors on or beside the road such as large trucks, buildings, trees, walls, overpasses or tunnels can result in a large variation in the number of received signals, and a large variation regarding the quality of the GNSS measurements and the independence of these errors both in time and across measurements. As an example, the well documented occurrence of Non Line-of-Sight (NLoS) tracking conditions will tend to create a GNSS pseudorange measurement error distribution with a heavy positive tail. This same situation might, if the obstacle at the origin of the NLOS situation is large enough to affect several satellites, result in correlating the error of multiple measurements. These phenomena are not necessarily well handled by traditional integrity mechanisms because they create breaches of fundamental assumptions (for instance ability to perform CDF overbounding or the assumption of measurement independence). This might result in an under-estimation of the targeted rate of Hazardous Misleading Information (HMI) required by the application. u-blox is actively working on various integrity solutions that would allow to be resistant to the above phenomenon. These solutions exploit differently the characteristics of the GNSS measurement errors, typically through the use of a strict measurement selection process and their exploitation by either a sequential filter (to capitalize on time filtering) or an advanced snapshot filter (taking advantage of the fine modeling of the GNSS measurement errors). Of course, the combined effect of strict measurement selection and sky-view obstruction can create frequent situations where a GNSS-only navigation system with integrity becomes unavailable. In this case, it appears critical to be able to use additional and complementary sensors. The typical sensors available on cars are IMUs and odometers, and these sensors are often used to improve accuracy and availability. However, their use does not necessarily conform to stringent integrity requirements, particularly ensuring that the required HMI rate is maintained. The objective of this paper is to present how the IMU and WT measurements are integrated in both the sequential and snapshot integrity solutions. Both methods require distinct solutions that allow for different advantages and pose constraints. The key element is to ensure that the solution remains appropriately bounded in various situations that range from the availability of a large number of good GNSS measurements to a reasonably long GNSS outage. For a sequential integrity algorithm, sensor data fits naturally into the estimation framework. The major challenges are in ensuring that the sensor models (both deterministic and stochastic) are adequate to maintain valid bounds, and that the non-linearity in propagation is handled properly. For the snapshot integrity algorithm, there is no built-in notion of bound propagation, and this concept must be explicitly handled by a new algorithm. When bound propagation is used to reduce latency or computational load, one option is to use GNSS delta-phase observations. During GNSS outages, this is clearly not possible, and IMU and WT sensors are the only available option. In this case, the propagation algorithm is somewhat similar to that used in the EKF, though it requires special care. Results presenting the bounding capability of both the sequential and snapshot solutions will be shown on an extensive set of more than 100 hours of real data collected in a large number of environments going from city centers to open sky roads. The performance in terms of protection level magnitude as a function of the “GNSS environment” will be presented as well as the comparison between both integrity solutions. Conclusions and way forward will then be formulated.

This paper describes a novel automotive snapshot integrity algorithm for bounding position, based on modelling GNSS measurements with non-Gaussian error distributions. A Bayesian method is used to derive the posterior probability distribution on position given a set of pseudorange and carrier phase observations from a single epoch. MCMC is then used to obtain rigorous probabilistic bounds on position. The MCMC method uses a novel form of parallel tempering to properly sample the multimodal posterior distribution created by carrier phase integer ambiguities, and importance sampling to obtain faster than real-time computational performance. Experimental results based on 27 hours of road driving show that integrity is maintained properly, with bounds which are significantly tighter than a more conventional EKF approach.

For several years now, various market studies have highlighted the importance of the delivery of positioning or navigation information with a high level of integrity for a wide range of use cases, the first of which being land and air vehicles. This feature is, for instance, a key enabler for the highly anticipated autonomous vehicle application. Although aviation has succeeded in reaching significant levels of GNSS-based integrity performance when the aircraft is in the air, the challenges to do so for vehicles mostly operating in or near urban areas are still considerable and include: - Lack of well-defined performance requirements - The GNSS propagation channel being considerably degraded compared to an open environment, for example, due to the presence of signal blockage, strong multipath, NLOS situation, interference, etc - The potential difficulty to use civil aviation ground or satellite augmentation systems including: no worldwide system, blocked geostationary satellites in cities, systems tuned for aviation requirements, etc - Use of GNSS carrier phase measurements and complementary navigation sensors to maintain a high availability and accuracy. Problems include the addition of new failure modes and the complex measurement characterization for various sensors - Use of fusion filters that rely on strong input assumptions, approximate propagation models and are sensitive to the influence of time-correlated measurement errors - Constraints on the computational and memory resources as well as on the platform price for mass market solutions - The complexity of verifying and certifying the performance of a navigation platform in all targeted environments u-blox has dedicated a significant amount of effort in the recent years to be able to respond to these challenges. The paper will introduce some of u-blox’s activities towards tackling the issue of providing navigation information with high integrity in degraded environments. This paper will specifically focus on the automotive case and will be centered on the following elements: - Platform overview: this section will briefly introduce the current u-blox platform for Lane Accurate Positioning and its general accuracy and availability performance -Fault characterization and integrity performance requirements providing: a) an overview of the different sensors and sensor faults characteristics b) the performance requirements that are targeted for Lane Accurate Positioning as well as elements of the Fault Tree Analysis - Accurate modeling of GNSS measurements: this section will discuss the methodology used to accurately model tracking errors impacting GNSS measurements in urban and sub-urban environments. This effort is currently based on a massive data collection effort associated with advanced modeling mechanisms: a) strategy to collect a representative and statistically meaningful set of data b) ability to evaluate the true measurement errors c) ability to classify the reception conditions based on multiple quality indicators d) building of measurement models fit for integrity monitoring - Pre-processing mechanisms: With the fine knowledge of the measurement models, various pre-processing (or measurement selections) mechanisms can be put in place. This has the great advantage of limiting the amount of non-nominal/faulty measurements entering the PVT filter. This selection process can be done very efficiently in a multi-GNSS and multi-sensor context and can have a significant impact on the design of the integrity monitoring mechanism that will protect the PVT solution. - Multi-sensor integrity monitoring considerations: this section will provide an overview of the integrity monitoring mechanisms that are promising, including in a multi-sensor context. The constraints related to the platform computational and memory resources will also be discussed. - Example of integrity performances in challenging environments: the last section will provide first promising results based on real data collected in various environments, cities and countries. Innovation: The significance of the presented work lies in the demonstration of the concrete and promising steps taken by a leading mass market chip manufacturer towards the design of an integrity receiver for the automotive market.

A magnetometer is often used to aid heading estimation of a low-cost Inertial Pedestrian Navigation System (IPNS) without which the latter will not be able to accurately estimate heading for more than a few seconds, even with the help of Zero Velocity Update (ZVU). Heading measurements from the magnetometer are typically integrated with gyro heading in an estimation filter such as Kalman Filter (KF) — to best estimate the true IPNS heading, resulting in a better positioning accuracy. However indoors the reliability of these measurements is often questionable because of the magnetic disturbances that can disrupt the measurements. To solve this problem, a filtering method is often used to select the best measurements. However, the importance of the frequency of these measurement updates has not been highlighted.

Collaborative (or cooperative) positioning or navigation uses multiple location sensors with different accuracy on different platforms for sharing of their absolute and relative localizations. Typical application scenarios are dismounted soldiers, swarms of UAV's, team of robots, emergency crews and first responders. This paper studies the challenges to realize a public and low-cost solution, based on mass users of multiple-sensor platforms. For the investigation field experiments revolved around the concept of collaborative navigation in a week at the University of Nottingham in May 2012. Different sensor platforms have been fitted with similar type of sensors, such as geodetic and low-cost high-sensitivity GNSS receivers, tactical grade IMU's, MEMS-based IMU's, miscellaneous sensors, including magnetometers, barometric pressure and step sensors, as well as image sensors, such as digital cameras and Flash LiDAR, and ultra-wide band (UWB) receivers. The employed platforms in the tests include a train on a building roof, mobile mapping vans and personal navigators. The presented preliminary results of the field experiments show that a positioning accuracy on the few meter level can be achieved for the navigation of the different platforms.

Foot mounted inertial sensors have been used in recent literature to provide high accuracy positioning through the use of zero velocity updates (ZUPT) every time the user takes a step. When only ZUPTs are used, the remaining positioning errors are primarily a result of heading drift due to poor observability.This paper demonstrates that a single axis rotation of an Inertial Measurement Unit (IMU) provides improved observability of IMU accelerometer and gyro biases. In particular, all gyro biases become observable when the IMU is rotated about a single horizontal axis. This results in a significant reduction of heading errors and hence also improves positioning accuracy.This paper first presents results using simulated data from a static environment to verify observability of bias states. The paper then describes the results from a physical implementation of a rotating IMU platform that is attached to a user's shoe. It is demonstrated that maximum position errors are significantly reduced to less than 1.3m over three trials for the rotating IMU compared to 12.4m for the non-rotating IMU.

These days, we seem to be breeding generations of people who are incapable of the simplest of tasks without computer/electronics assistance. Whilst most people find it hard to understand the simplicity of maps anymore, it is not a surprise when more and more people opt to follow their trustiest companion i.e. Global Positioning System (GPS) for turn-by-turn directions with mindless devotion and blind faith. “No matter what kind of GPS you have, relying solely on it is a bad bet.” Mohd Yahya, a PhD candidate in Engineering Surveying and Space Geodesy at the University of Nottingham said, “GPS is a radio-navigation system and is often limited by signal interference and signal outage particularly in urban environment with high-rise buildings, trees, tunnels and underground motorways.” He added, “In light of urgent need for robust and sustained navigation approach, our research team at the Nottingham Geospatial Building is investigating a new and innovative technique called Peer-to-peer (P2P) collaborative positioning to further improve GPS navigation capability of a group of users.” This P2P approach benefits from a joint position solution obtained via a network of users who may be able to receive sufficient satellite signals, augmented by inter-users ranging measurements and information exchange. As P2P approach allows neighbouring users to improve one’s position by communicating and sharing positioning information with each other, it is expected that the proposed intelligent positioning approach is capable in improving safety, efficiency and accessibility of transit and highway travel worldwide. (End of Press Release- UoN).

Low cost inertial sensors are one potential method for positioning indoors. However, such sensors typically provide poor quality measurements which are only suitable for positioning for a few seconds. For inertial navigation to be useful, it is necessary to combine the sensors with measurements from other systems.This paper explores the integration of an Inertial Measurement Unit (IMU) with measurements from a computer vision algorithm, for indoor pedestrian navigation.. The concept is to make use of sensors that are already available in modern smartphones. It is assumed that a pedestrian user is walking with the mobile device held out in front of them with the camera pointing approximately towards the ground. Therefore the camera is taking images of the ground immediately in front of the user.The computer vision algorithm matches features between pairs of successive images. Typically, many of these features will fall on the ground plane. The relative positions of features on the ground plane are related by a homography which describes the rotation and translation of the camera between images. The robust BaySAC framework is used to simultaneously identify which features lie on the ground plane, while estimating the homography relating the two views. From the homography, the camera's orientation and 3-dimensional body frame translation relative to its previous position are computed. This information, along with measurements from other systems such as GPS when they are available, is used to aid the IMU using a Kalman filter, to reduce the position drift.This paper describes the implementation of the combined computer vision and inertial navigation approach. A microelectromechanical (MEMS) IMU is used along with a consumer grade digital camera to capture data. It is demonstrated that the drift of the inertial sensor is significantly reduced by incorporating measurements from the computer vision algorithm. The algorithm is relatively computationally expensive, therefore this paper explores the computational requirements and identifies two methods that may be used to improve efficiency. A method of reducing the sample rate of the computer vision algorithm is demonstrated to provide a significant reduction in processor requirements with only a small reduction in positioning accuracy.

This paper describes a scheme for pedestrian navigation integrating measurements from a foot-mounted IMU with position and orientation updates from computer vision techniques. By mounting an IMU on a user's foot, the position drift can be substantially reduced since zero velocity updates can be applied every step. However, such a system will still suffer from position drift unless occasional measurements are available from other sensors. This paper describes a novel method for restricting such position drift using an image recognition algorithm. Firstly, a database of images and their locations is constructed over an area of interest. A user then navigates the area using foot-mounted inertial sensors and a video camera. As images are acquired, they are used to search the database of images using the Image Bag-of-Words algorithm. When new images are successfully matched with images in the database, the position from the database is used to update the inertial position using a Kalman filter. Furthermore, when images are successfully matched, orientation updates can be applied by estimating the relative orientation of the two cameras. These measurements can help overcome the limitations of the IMU, in particular the problem with heading drift. The integrated inertial and vision system is demonstrated to pro-vide better than 10m accuracy (typically 1-5m) over a period of 21 minutes, and the paper demonstrates how orientation updates could be applied in the future.

Low cost inertial sensors are often promoted as the solution to indoor navigation. However, in reality, the quality of the measurements is poor, and as a result, the sensors can only be used to navigate for a few seconds at a time before the drift becomes too large to be useful. Therefore, it is necessary to regularly update the sensors with measurements from external systems such as GPS or other sensors useful for navigation. One such sensor is provided by the computer vision community where a camera can be used to obtain information about the relative translation and rotation between successive images. This paper describes the use of a camera attached to a low cost IMU for navigation in areas where GPS is unavailable such as indoors or deep urban canyons. It is assumed that a pedestrian user is walking with the mobile device held out in front of them with the camera pointing approximately towards the ground. Features are matched between successive frames, and the robust RANSAC framework is used to identify which of these lie on the ground plane, while estimating the camera’s orientation and 3 dimensional body frame translation relative to its previous position. This information is used to aid the IMU using a Kalman filter to reduce the position drift. This paper describes the implementation of the combined computer vision and inertial navigation approach. A tactical grade IMU is used for initial testing since it provides more reliable measurements and enables us to provide a reference by which to compare the measurements obtained from the computer vision algorithm. It is demonstrated that even with a good quality IMU, the algorithm is able to significantly improve the performance of INS navigation when GPS measurements are unavailable.

The use of orthomosaic images from aerial or satellite data are increasingly common. While current acquisition methods are cost-effective on a national or regional scale, local scale imagery is prohibitively expensive for many target applications. In this paper we present a combined hardware and software solution, developed at the Geospatial Research Centre, which aims to reduce the cost of acquiring and processing imagery and related data in order to produce orthomosaics in a cost-effective manner on a small, local scale.The hardware component consists of a combined GNSS and inertial solution for determining the position and orientation of a sensor, typically a consumer-grade camera such as a digital SLR. The combination of imagery and navigation metadata allows images to be directly geo-referenced by projecting them on to readily available surface models. Refinements to this initial processing are also presented, which account for boresight and lens calibration error; automatically establishing a correspondence between image features for bundle adjustment; and reducing the visual appearance of any residual misalignments in the final mosaic. The use of commodity sensors and automated processing is an important step in reducing the cost of image acquisition and orthomosaic generation.The methods described are illustrated using two sample sequences. The first is a set of visible images captured from a digital SLR, and the second a set of frames extracted from a thermal video sequence. These two sequences demonstrate the range of imagery that can be processed, which can support applications ranging from environmental monitoring and precision agriculture to urban planning and infrastructure maintenance.

Foot mounted inertial sensors provide a promising method for accurate pedestrian positioning. By mounting sensors on a user's foot, the accumulated position drift from dead reckoning can be reduced by applying zero velocity updates using a Kalman filter every time a user takes a step. However, such a system will still suffer from position drift unless occasional position updates are available. This paper describes a novel method for restricting such position drift using an image recognition algorithm. Firstly, a database of images and their locations is constructed over an area of interest. A user then navigates the area using foot-mounted inertial sensors and a video camera. As images are acquired, they are used to search the database of images using the Image Bag-of-Words algorithm. When new images are success-fully matched with images in the database, the position from the database is used to update the inertial position using a Kalman filter. GNSS updates can also be used in the filter when available. The integrated inertial and vision system is demonstrated to provide better than 10m accuracy (typically 1-5m) over a period of 21 minutes. The system is relatively inexpensive, could run in real-time, does not require costly infrastructure, and could be deployed over larger areas.

The Institute of Engineering Surveying and Space Geodesy (IESSG), in collaboration with five other universities and industrial partners in the UK, is involved in two major projects concerned with locating and positioning buried assets in a non-invasive manner. GNSS based technology can be used to map these assets, however, most of the environments are built up areas where GNSS signals are lost, unavailable or have large errors. The aim of the projects is to tackle the issue of positioning in built up areas in two stages. The first stage is to integrate GPS with other sensors such as INS and a total station. The second stage is to analyse using simulation, the impact that future GNSS developments will have when there are three fully operational satellite navigation systems, i.e. GPS, Galileo and GLONASS. This paper gives an overview of these projects, illustrating the research being developed in each of the above areas. Preliminarily results using a Leica SmartStation and GPS integrated with INS are presented with analysis and discussion of the results.

GPS and low-cost INS sensors are widely used for positioning and attitude determination applications. Low-cost inertial sensors exhibit large errors that can be compensated using position and velocity updates from GPS. Combining both sensors using a Kalman filter provides high-accuracy, real-time navigation. A conventional Kalman filter relies on the correct definition of the measurement and process noise matrices, which are generally defined a priori and remain fixed throughout the processing run. Adaptive Kalman filtering techniques use the residual sequences to adapt the stochastic properties of the filter on line to correspond to the temporal dependence of the errors involved. This paper examines the use of three adaptive filtering techniques. These are artificially scaling the predicted Kalman filter co-variance, the Adaptive Kalman Filter and Multiple Model Adaptive Estimation. The algorithms are tested with the GPS and inertial data simulation software. A trajectory taken from a real marine trial is used to test the dynamic alignment of the inertial sensor errors. Results show that on line estimation of the stochastic properties of the inertial system can significantly improve the speed of the dynamic alignment and potentially improve the overall navigation accuracy and integrity.

GPS and Inertial Navigation Systems are used for positioning and attitude determination in a wide range of applications. Over the last few years, a number of low cost inertial sensors have become available. Although they exhibit large errors, GPS measurements can be used correct the INS and sensor errors to provide high accuracy real-time navigation. The integration of GPS and INS measurements is usually achieved using a Kalman filter. The measurement and process noise matrices used in the Kalman filter represent the stochastic properties of the GPS and INS systems respectively. Traditionally they are defined a priori and remain constant throughout a processing run. In reality, the stochastic properties of the system vary depending on factors such as vehicle dynamics and environmental conditions. This is particularly an issue for low cost inertial sensors where the initial sensor errors can be large, and experience significant temporal variation. This paper investigates three adaptive Kalman filtering algorithms that can be used to improve the estimation of the stochastic properties of a low cost INS. The algorithms are tested using a low cost Crossbow MEMS IMU integrated with carrier phase GPS for a marine application. The adaptive Kalman filtering algorithms are shown to reduce the dependence on the a priori information used in the filter. This results in a reduction in the time required to initialise the sensor errors and align the INS, and results in an improvement in navigation performance.