Active safety systems for automotive vehicles rely on estimates of the unmeasured vehicle velocity vector. This velocity vector is easily estimated using kinematic equations if the wheels do not slip on the road, but it is precisely in situations involving significant wheel slip that an accurate estimate of the velocity vector is most needed. This paper presents a velocity estimation algorithm for automotive vehicles that exploits dynamic equations to avoid any use of zero wheel slip assumptions in its implementation, and it requires only those sensors already installed in vehicles equipped with active safety systems. The proposed estimation algorithm is based on a nonlinear least-squares formulation for approximate inversion of the system's state-to-measurement map.