This paper focuses on the development and experimental evaluation of a novel adaptive feedforward vibration cancellation based friction estimation system. The friction estimation utilizes a small instrumented redundant wheel on the vehicle. Unlike other systems previously documented in literature, the developed system can provide a continuous measurement of the friction coefficient under all vehicle maneuvers, even when the longitudinal and lateral accelerations are both zero. A key challenge in the development of the estimation system is the need to remove the influence of vibrations and the influence of vehicle maneuvers from the measured signal of a force sensor. An adaptive feedforward algorithm based on the use of accelerometer signals as reference inputs is developed. The parameters of the feedforward model estimated by the adaptive algorithm themselves serve to determine the value of the friction coefficient. Detailed experimental results are presented on a skid pad wherein the road surface changes from dry asphalt to ice.