Ifsttar PhD subject

French version
Detailed form :

Title : Analysis of the progress due to GNSS new constellations, signals and high accuracy services, as well as 3D high definition maps, in urban geolocalisation

Main host Laboratory - Referent Advisor COSYS - SII  -  BETAILLE David      tél. : +33 240845623

Director of the main host Laboratory DUMOULIN Jean  -

PhD Speciality Automatique et traitement du signal

Axis of the performance contract 1 - COP2017 - Efficient transport and safe travel

Main location Nantes

Doctoral affiliation UNIVERSITE GUSTAVE EIFFEL

PhD school Mathématiques et Sciences et Technologies du numérique, de l'Information et de la Communication (MaSTIC)

Planned PhD supervisor BETAILLE David  -  Université Gustave Eiffel  -  COSYS - SII

Planned financing Thèse sur contrat  - Université Gustave Eiffel

Abstract

Applications of GNSS satellite positioning systems are becoming more common in land transport systems, whether road or rail. Future applications, however, require increasingly demanding performance in terms of accuracy as well as integrity. This is particularly the case in the development of autonomous vehicles (cars or trains) or highly critical systems for safety such as railway signaling.
These systems are also used in complex environments for the reception of GNSS signals such as urban environments or railway trenches, in which local phenomena present a significant source of error.
The work of [Zhu 2018] has shown that the combination of error models adapted to local phenomena and fault detection and exclusion techniques (FDE) can both reduce the inaccuracy of the estimated positions but also improve the integrity of the GPS solution by ensuring the exclusion of unacceptable residual errors.
With the deployment of new constellations (especially Galileo) and new signals (those more robust to multipath with Galileo in particular), and in the following of the aforementioned doctoral thesis, it is necessary to explore the potential of these solutions for multi-channel and multi-constellation systems. The thesis work proposed here consists, first of all, in adapting the algorithms to these new configurations and in evaluating their performance after having carried out a measurement campaign and constituted a test database.
The satellite positioning techniques targeted will use the phase of the carrier in PPP (Precise Point Positioning), that is to say in autonomy. The most accurate RTK technique today relies on interferometry between two nearby antennas (a few kilometers). Atmospheric disturbances are the same for the base and the rover. In PPP, there is no basis. But a network of bases, combined, providing corrections for any rover in the area. PPP does not give a centimeter solution in seconds like RTK. There is a more or less long convergence time. On going works in H2020 eMaps project will be used.

In a second step, we propose to address the contribution of the map to these solutions. In the works of [Zhu, 2018] a 3D map was used to detect the reception state of the received signals on the one hand and to estimate the delay caused by NLOS reception by ray tracing on the other hand in order to correct it. These detection and correction were made based on the knowledge of the true position in the map, that is to say provided by the reference trajectometry system (an inertial unit coupled to GNSS PPK), which is not available a priori. These calculations (ray tracing) must be done from an approximate position, which can be inferred in real time by Kalman filtering. This position can also be projected (map-matched at lane level) on different assumptions of taxiways [Bétaille, 2017].
In this new thesis, we want to explore the potential of a map-matched solution to improve and simplify the use of the map for the ray tracing stage.
Finally, we will set the framework for a concept of monitoring the integrity of a map-matched position in order to provide the user with a set of information {Id_route; Id_lane; ATPL}, ie: identifiers of the road and taxiway on multi-lane road, associated with a longitudinal protection level (along track) characterizing the location error bounds on this road.

In conclusion, the contribution of the 3D map has been shown so far under the assumption that the positioning precision is enough to properly search for nearby buildings in this map. It’s actually a chicken and egg problem. In fact, most of the published work uses the ground truth position to read the map properly. (This “true” position being obtained, for example, by sophisticated equipment, such as an inertial unit.) However, in practice, this is not feasible.

Therefore one must provide oneself with a means of positioning that is precise enough to read the map, and not too expensive for the automotive market. Hence the originality in this thesis to take advantage of all constellations, not just GPS, and all frequencies, not just L1. This thesis could, if the results show a satisfactory proof of concept, interest manufacturers like Ublox for future valuation.

Finally, in addition to precision, it is also integrity, that is to say the ability of the algorithm to properly estimate the error it makes, that we want to investigate. Again, much of the past work is showing results, but there is no evidence that it is indeed the 3D map that improves integrity. Because the improvement proposals are mixed there and one never knows what is really due to the map, all other things being equal.

Keywords : Keywords : Geopositioning, GNSS, Quality of Service, integrity, sensor data fusion, Bayesian estimation, digital maps, map-matching, lane-level positioning

Main host Laboratory - Referent Advisor	COSYS - SII - BETAILLE David tél. : +33 240845623
Director of the main host Laboratory	DUMOULIN Jean -
PhD Speciality	Automatique et traitement du signal
Axis of the performance contract	1 - COP2017 - Efficient transport and safe travel
Main location	Nantes
Doctoral affiliation	UNIVERSITE GUSTAVE EIFFEL
PhD school	Mathématiques et Sciences et Technologies du numérique, de l'Information et de la Communication (MaSTIC)
Planned PhD supervisor	BETAILLE David - Université Gustave Eiffel - COSYS - SII
Planned financing	Thèse sur contrat - Université Gustave Eiffel
Abstract Applications of GNSS satellite positioning systems are becoming more common in land transport systems, whether road or rail. Future applications, however, require increasingly demanding performance in terms of accuracy as well as integrity. This is particularly the case in the development of autonomous vehicles (cars or trains) or highly critical systems for safety such as railway signaling. These systems are also used in complex environments for the reception of GNSS signals such as urban environments or railway trenches, in which local phenomena present a significant source of error. The work of [Zhu 2018] has shown that the combination of error models adapted to local phenomena and fault detection and exclusion techniques (FDE) can both reduce the inaccuracy of the estimated positions but also improve the integrity of the GPS solution by ensuring the exclusion of unacceptable residual errors. With the deployment of new constellations (especially Galileo) and new signals (those more robust to multipath with Galileo in particular), and in the following of the aforementioned doctoral thesis, it is necessary to explore the potential of these solutions for multi-channel and multi-constellation systems. The thesis work proposed here consists, first of all, in adapting the algorithms to these new configurations and in evaluating their performance after having carried out a measurement campaign and constituted a test database. The satellite positioning techniques targeted will use the phase of the carrier in PPP (Precise Point Positioning), that is to say in autonomy. The most accurate RTK technique today relies on interferometry between two nearby antennas (a few kilometers). Atmospheric disturbances are the same for the base and the rover. In PPP, there is no basis. But a network of bases, combined, providing corrections for any rover in the area. PPP does not give a centimeter solution in seconds like RTK. There is a more or less long convergence time. On going works in H2020 eMaps project will be used. In a second step, we propose to address the contribution of the map to these solutions. In the works of [Zhu, 2018] a 3D map was used to detect the reception state of the received signals on the one hand and to estimate the delay caused by NLOS reception by ray tracing on the other hand in order to correct it. These detection and correction were made based on the knowledge of the true position in the map, that is to say provided by the reference trajectometry system (an inertial unit coupled to GNSS PPK), which is not available a priori. These calculations (ray tracing) must be done from an approximate position, which can be inferred in real time by Kalman filtering. This position can also be projected (map-matched at lane level) on different assumptions of taxiways [Bétaille, 2017]. In this new thesis, we want to explore the potential of a map-matched solution to improve and simplify the use of the map for the ray tracing stage. Finally, we will set the framework for a concept of monitoring the integrity of a map-matched position in order to provide the user with a set of information {Id_route; Id_lane; ATPL}, ie: identifiers of the road and taxiway on multi-lane road, associated with a longitudinal protection level (along track) characterizing the location error bounds on this road. In conclusion, the contribution of the 3D map has been shown so far under the assumption that the positioning precision is enough to properly search for nearby buildings in this map. It’s actually a chicken and egg problem. In fact, most of the published work uses the ground truth position to read the map properly. (This “true” position being obtained, for example, by sophisticated equipment, such as an inertial unit.) However, in practice, this is not feasible. Therefore one must provide oneself with a means of positioning that is precise enough to read the map, and not too expensive for the automotive market. Hence the originality in this thesis to take advantage of all constellations, not just GPS, and all frequencies, not just L1. This thesis could, if the results show a satisfactory proof of concept, interest manufacturers like Ublox for future valuation. Finally, in addition to precision, it is also integrity, that is to say the ability of the algorithm to properly estimate the error it makes, that we want to investigate. Again, much of the past work is showing results, but there is no evidence that it is indeed the 3D map that improves integrity. Because the improvement proposals are mixed there and one never knows what is really due to the map, all other things being equal.
Keywords :	Keywords : Geopositioning, GNSS, Quality of Service, integrity, sensor data fusion, Bayesian estimation, digital maps, map-matching, lane-level positioning

List of topics

Applications closed