Accurate information about vehicle longitudinal and lateral velocities is vital for efficient operation of many vehicle control systems. In this paper, an estimation structure to simultaneously estimate vehicle velocities in longitudinal and lateral directions is developed and experimentally validated. This structure includes two parallel estimators: the first estimator is a kinematic-based observer for longitudinal velocity estimation, and the second is a combination of a kinematic-based observer and an inverse tire model to estimate vehicle lateral velocity. The proposed structure can effectively handle the additive biases, which are common in vehicle's stock accelerometers' signals, and provide accurate estimate of vehicle velocities when one (or more than one) wheel experiences the excessive slippage. Additionally, the proposed structure is not sensitive to changes in parameters of tire model and vehicle mass. The performance of this estimation structure is validated by experimental studies. A 0018-9545 (c)