The subject of this article is about study and realisation of a tight hybrid filter for vehicle positioning in dense urban area. In our approach we fused the measurements coming from the sensors in a hybrid filter at the pseudo-ranges level. The proposed filter combines the "observables" (measurements) of phase and code of a GPS receiver with the measurements of an odometer and a gyrometer. We describe in this article the design of the fusion filter inspired of the distributed "track to track" Kalman filter architecture. In our approach the pseudoanges rspredictions, construct with the phase measurements of the GPS and the Dead-reckoning measurements system, are fused. The fused predictions are them corrected with the GPS code measurements of the pseudo-ranges. We use the classical snapshot least squares estimation technique to calculate the position with the estimated pseudo-ranges. We present in this article different simulations results obtained in a real dense urban context. We show that the tight hybrid filter we propose gives a location at least as precise as the classical loosely coupled hybrid filter. This result is obtained for a number of satellites lower than four and various values of the HDOP (Horizontal Dilution Off Precision).