Accurate information of the absolute speed of a vehicle, when available, can be vital in simplifying the control laws of an anti-lock braking system (ABS) and auto-traction system (ATS). A current meter for measuring the speed of a vehicle is to multiply the measured wheel rotation rate to the wheel radius. The approach often includes abrupt unpredictable errors due to slip and skid of wheels and a biased error due to the steady state slip. These errors are sources of difficulty in the implementation of an ABS that is based on the absolute speed of the vehicle.This paper describes an accurate rule-based Kalman filtering technique for estimating the absolute speed of a vehicle. The enhanced accuracy is achieved by employing an additional accelerometer to complement the wheel speed-based speedometer. The accelerometer measures the acceleration of the vehicle in its forward direction and may be corrupted as well by high frequency noise.The proposed sensing technique employs the extended Kalman filter to reduce the high frequency noises in the acceleration measurement and the biased error coming from the wheel speed measurement. Furthermore, we incorporate a rule-based strategy that switches the values of the Kalman filter coefficients to compensate for abrupt nonstationary errors in the measurement. The rules are developed based on a simple set of knowledge. For example, when skid or slip occurs, the wheel speed measurement may be quite erroneous; therefore, heavier reliance on the acceleration measurements would be placed over that of the wheel speed. Simulations and laboratory model experiments were carried out to verify the proposed rule-based estimation strategy and prove that it can accurately estimate the absolute speed even when wheel slip or skid occurs.