In this paper, a modified adaptive nonlinear filter (MANF) to estimate the vehicle longitudinal velocity of the vehicle has been presented which has an acceptable accuracy and simple structure. Obtaining accurate model with minimum uncertainty is difficult and almost impossible, consequently, conventional model based filters can not be usded simply in estimating the longituadinal velocity of the vehicle during the braking. On the other hand, the model based filters reuire the braking toque in velocity estimation where, measuring the braking torque requires specific sensors which will cause increasing the costs of repairing and maintenance. Consequently,  in this paper a MANF has been introduced which didn’t require the dynamic model of vehicle and the value of braking torque in estimation process. MANF has been evaluated by experimental data and its effectiveness has been approved.