Vehicle's energy estimation using low frequency speed signal.

Everyday vehicles fleets positions and speeds are collected by private companies and public authorities, stored in servers and used for vehicles tracking and diagnostic. These data are usually sampled and transferred over the long-range wireless network with low sampling rates for cost savings in term of network bandwidth and storage. This policy limits the use of the data for fine-grained estimation of vehicle energy, and therefore of fuel consumption and CO2 emissions. This paper analyzes the effect of a low-frequency sampling rate of inertial variables on vehicle’s energy estimation. It statistically quantifies the error with respect to accurate signals. An experimental campaign carried out on different vehicles in different days supports the analysis. Experimental results show that the estimation error mean and variance depends on the sample rate and the time window size and that they do not significantly varies within vehicles of the same class.