Localización de vehículos en entornos urbanos mediante GPS y mapas 3D

  1. Carolina de los Ángeles Piñana Díaz
Supervised by:
  1. Rafael Toledo Moreo Director
  2. Antonio Skarmeta Gómez Director

Defence university: Universidad de Murcia

Fecha de defensa: 14 July 2017

  1. Pedro J. Navarro Lorente Chair
  2. José Santa Lozano Secretary
  3. Fernando Daniel Quesada Pereira Committee member

Type: Thesis


1. Introduction. Many applications based on vehicle localization, such as navigation systems, fleet management or Electronic Toll Collection (ETC), are a reality today thanks to the so-called Global Navigation Satellite Systems (GNSS). All of them require an accurate position of the receiver. However, location-based applications must face serious drawbacks in certain scenarios such as tunnels, covered parking lots and urban canyons. Beyond the lack of coverage, GNSS positioning suffers from several sources of error, such as those caused by clocks and satellite orbits, or perturbations in the ionosphere and the hydrosphere. Among all these errors the most complex to model and compensate is the so-called multipath error. This error occurs due to the reflections of the satellite signals on flat surfaces, such as buildings. Given that the path travelled by a reflected signal is always longer than the path travelled by a direct signal, multipath affected signals are delayed, causing GPS outliers and wrong estimates of the position. Consequently, their estimated pseudorange values are not valid and the final estimate of the vehicle positioning is erroneous because of the inaccuracies in the trilateralization process. This thesis describes a method to distinguish direct signals from satellites from those who arrive as a result of successive rebounds on the surfaces of the nearest buildings. 2. Objectives. The main objectives of this work are the following: 1. To develop an EEMap that stores the most relevant parameters of the nearest buildings to the receiver position and helps to solve the multipath problem. 2. To develop a visibility algorithm to discard satellite signals affected by multipath. 3. To implement a DOP (Dilution Of Precision) algorithm to estimate the quality of the satellite geometry. 4. To develop a positioning algorithm for the calculation of the receiver position using the pseudoranges measurements from satellite signals in direct view. 5. To integrate the elevation models with the visibility algorithm and with the positioning algorithm to obtain the estimation of the position of the receiver. 6. To integrate the systems described above to determine the error-free position. 3. Methodology. The first step is the modeling of the environment. For this purpose, the concept of EEMap (Elevation Enhanced Map) is defined. An EEMap is a 3D digital map that stores relevant information about the scenario to be modeled. The process of creation of the EEMap is described in this thesis. Once the map is constructed, a visibility detection algorithm is developed. This algorithm allows to discard those satellites signals affected by multipath and to eliminate wrong pseudoranges. Then, the receiver position is recalculated using only data from satellites which are in direct view. Two algorithms are proposed to calculate the position of the receiver from the pseudodistances of the satellites. The first one is called "Least Square Algorithm" (LSQ) and the second one is the "Bancroft Algorithm". 4. Results. In order to verify the validity of the described concepts, several test campaigns have been carried out in real scenarios. Data were collected in urbanized areas of Spain and France with high buildings and narrow streets with limited visibility. The results of these tests are presented, analyzed and compared with the position provided directly by the GPS receiver itself. Among other measurements, statistical calculations of the DOP (Precision Dilution) and HPE (Horizontal Positioning Error) parameters are presented. The correctly identified NLOS (Non Line Of Sight) detection rates are closed to 100% and some false positives are obtained, with some small variations depending on each test. Even though the discard of spurious measurements leads to lower position availability, the reliability of the solution is certainly far superior. 5. Conclusions. Overall, this thesis presents two major contributions to the field of road vehicle positioning. Both of them have been published in recognized magazines of the sector. Firstly, the EEMaps proposal is sufficiently complete to model buildings in an accurate way to deal with the proposed problem. Maps do not need to handle heavy data, and do not require costly computation. The accuracy of the proposed EEMaps has been validated by means of an international reference on 3D maps. Secondly, these maps have shown that they can effectively support satellite navigation. They can serve as an input to a GNSS satellite detection algorithm to eliminate signals from satellites which are not in direct view with the receiver.