The new location-based automotive applications require the introduction of the integrity concept, as developed in the aviation domain. This thesis aims at improving the localization accuracy and at developing integrity mechanisms for mass-market ground vehicles. It focuses on urban environments. The aeronautical approach makes only use of GPS measurements and comprises two steps: the fault detection and exclusion (FDE), then the protection level computation which defines a safety area. However, some experiments have pointed that urban environments provide little redundancy and multiple faults. This reduces heavily the FDE step efficiency. This thesis develops two approaches from the following hypothesis: the introduction of new redundant measurements from proprioceptive sensors embedded in ground vehicles (odometers, yaw rate sensor, steering wheel angle...) counterbalances the problems from the urban environment. The first approach puts forward new FDE algorithms which take advantage of dynamic filtering and Doppler measurements. The estimated performances, through four study cases, reveal a significant improvement in urban conditions. The second approach uses the trajectory estimation over a time horizon. This formalism allows introducing dynamic data (successive positions, proprioceptive measurements) into the standard integrity algorithms, which are originally static. Protection levels computed this way are divided by 3.
