Commercial vehicles transport the majority of the inland freight in US and a significant number of passengers. They are large fuel consumers as they operate a large number of hours per day, pulling heavy loads. The increasing fuel price and the Green House Gas emission regulation have provided a strong impetus for new technologies capable of improving the commercial vehicle fuel economy. Among others, optimized powertrain control can improve the vehicle fuel economy, particularly if it is based on accurate information about the instantaneous load demand. Furthermore, model-based online vehicle parameter estimator is critical for implementation of an adaptive vehicle controller.While vehicle mass estimation has been successfully demonstrated, rolling resistance and aerodynamic drag estimation has not been fully explored yet. This paper examines this problem using model-based approach with a supervisory data extraction scheme. The estimator uses signals from the standard vehicle CAN bus and accelerometer sensor to determine vehicle rolling resistance and aerodynamic drag coefficient on-the-fly. A supervisor monitors vehicle motion and extracts data only during events with high signal to noise ratio. The algorithm was tested using experimental data acquired from real-world driving cycles. The estimator is validated with a coast-down test and shows a sufficient accuracy in predicting the road load.