Making the grade: The sensitivity of Class 8 truck fuel consumption and operational behavior predictions to road grade data quality

The fidelity of road grade data has a large impact on predictions of vehicle fuel consumption and operational behavior. This work addresses this issue for Class 8 trucks by comparing predicted fuel consumption and operation (shifting, engine torque/speed, and braking) of a single Class 8 truck simulated with grade data for the same corridor from different sources. The truth baseline road grade of best fidelity available with LiDAR (Light Detection and Ranging) was obtained previously. This paper compares road grade data of a highway in Indiana to the truth baseline from four other typical methods (i) utilizing GPS (Global Positioning System) to record horizontal position and vertical elevation, (ii) logging the pitch of a cost effective IMU (Inertial Measurement Unit), (iii) integrating the horizontal and vertical velocities of the same IMU, and (iv) a commercially available dataset (Comm). Comm grade data ( R 2 = 0.99) best matches the LiDAR reference over a 5432 m stretch of US 231 where high quality LiDAR data was available, followed in quality by the integrated IMU velocity road grade ( R 2 = 0.98). IMU pitch ( R 2 = 0.96) and GPS ( R 2 = 0.12) were the least effective methods of road grade measurement, with IMU pitch over-predicting fuel consumption by 17.4%. GPS under-predicted fuel consumption by 3.8% and produced an engine torque output dramatically different from the other four road grade measurement techniques. Limitations of the Comm dataset are shown, namely missing road grade (decreased point density) for up to 1 km spans on other sections of US 231, as well as for other Indiana corridors surveyed by the author. Vehicle simulations show that both the Comm data (where available and accurate) and integrated IMU road grade data result in fuel consumption predictions within 2.5% of those simulated with the truth reference grade data.