Post-doc 2005-2006: Analysis et Rejecting GPS Multi-Paths
Location: LAAS, 7 avenue du colonel Roche, 31077 Toulouse Cedex 4, France
Duration: 12-16 months
Salary: 2150 € / month
Director: André Monin - e-mail: monin@laas.fr

This proposal is made within the framework of the European projectEureka "Egnos-bus" n°3398 which gathers two companies(Navocap-Toulouse_France and Etra-Valencia Spain) and a researchlaboratory (LAAS-CNRS).
The goal of this project is to develop a Exploitation Assistance System of a bus
network in towns. In particular, this system needs to regularly deliverthe localization of each bus constituting the fleet, as well asinformation relating to the integrity of provided information(integrity, uncertainties…). The localization is carried out fromsystem GPS (Global Positioning System). The goal is to reach alocalization precision of about 5 meters. To reach this one, one willuse correction information from geostationary satellite EGNOS whichwill be transmitted by VHF to GPS receivers embedded on vehicles.

However, in urban environment, the presence of multi-paths (reflectionson buildings) and the masking of satellites (4 satellites are necessaryto localization) strongly decrease the performances of GPS. Those beinglocal cannot be taken into account by EGNOS correction. Work suggestedconsists in improving the precision of GPS receiver and providingindicators of integrity, this by taking into account these disturbances(multi-paths and masking) and other data relating to movement(primarily odometry).
With this intention, it will be first necessary to identify the stochastic processes which
describe the effects of masking and multi-paths to pseudo-distancemeasure delivered by GPS receiver. Indeed, it is not considered withinthis framework, for economic reasons, to consider the rough signalprocessing. This identification will be made on the basis of real data,collected according to a protocol which will have to be defined.
From this modeling, it will necessary to develop a localizationestimator taking into account the non-linear and non-Gaussian aspectsof the system and which integrates data on vehicle (odometer). Itshould be noted that some tests of feasibility were carried outsuccessfully by a master student by using a modeling based on Gaussiandensity mixture. Competences necessary for the good realization of thisproject are in stochastic processes modeling and non-linear filtering.An experience on GPS system will be certainly welcome.