Aerodynamic drag is an important contributor to vehicle energy consumption, especially in highway conditions. Hence, estimating on-line the aerodynamic drag, or equivalently its coefficient, is an interesting challenge in order to reduce the energy consumption of vehicles. However, real systems are characterized by noisy sensor measurements. Extended Kalman Filter (EKF) is a commonly used algorithm for parameter estimation due to its stochastic filtering properties and is based on a first order approximation of the system dynamics. Similarly, the Unscented Kalman Filter (UKF) has been proposed as an alternative to the EKF in the field of nonlinear filtering and has received great attention in parameter estimation. This paper presents these two observers EKF and UKF for online estimation of the aerodynamic drag coefficient, based on noisy sensor measurements of a vehicle velocity, powertrain wheel torque and a longitudinal dynamic vehicle model. The design of the estimators is described and the performances are assessed against real measurements. The robustness is evaluated with different spoiler positions and the optimal position corresponding to a minimum of the aerodynamic drag coefficient is confirmed. Two different masses are used to validate each estimator.