Real-time information of tire friction forces is invaluable for vehicle control systems, such as ABS and the electronic stability program (ESP), to achieve better stability and maneuverability. To estimate tire forces on-line, this paper proposes a robust tire force estimation algorithm which is able to identify the longitudinal and lateral tire forces of each individual wheel. In addition, the estimation results are robust w.r.t variations in vehicle parameters. The dependency between the longitudinal and lateral tire forces is explicitly taken into account by incorporating friction ellipses into the estimation algorithm. Simulations based on a 14-degree-of-freedom nonlinear vehicle model are conducted and the results are satisfactory, even in the presence of sudden changes of the road conditions and variations in vehicle parameters.