Nowadays, active safety has a significant place in the automotive field. Recent studies have shown that electronic driver assistance systems could be improved if the wheel-ground interface could be determined in real time from the driver?s requests. So, the aim of this paper is to present a real time estimation of the forces exerted between the wheel and the road, using neural networks. The originality of this study consists in using a limited number of low-cost sensors for the partial wrench estimation. For the moment, the study only considers the estimation of forces Fx, Fy and Fz for purely lateral excitations of the vehicle. The results obtained from the test car of the laboratory are presented, as well as the future improvements to be made.