TERRAIN CHARACTERIZATION AND ITS IMPACT ON THE ENERGY MANAGEMENT OF AN ELECTRIC VEHICLE

Abstract

ELECTRIC VEHICLES have gained great attention in everyday usage and especially in the industry since they offer a green solution to a variety of tasks. During their assignments, electric vehicles traverse through terrains with different characteristics and thus different terramechanic parameters, which affect the energy consumed and might lower their autonomy. In this work, we face the problem of characterizing the energy consumption of an electric vehicle as it transits thorough different terrains, under different controlled speeds. To this end, we use a non-invasive sensing system based on arti cial stereo-vision, for terrain characterization and classi cation; RTK localization for position and speed estimation, and a prepared sensory system to monitor the batteries according to the behaviour of the vehicle during the trials. The aim is to offer a characterization of the energy management while traversing different terrains for enhancing routing planning problem (RPP) solutions and monitor autonomy of the vehicle. The system was tested in 4 different terrains at different speeds, showing a clear relation between the speed of the vehicle and the instantaneous power consumption (IPC) of the engine, we also noticed clear differences in the data depending on if the vehicle was speeding up or reached a constant speed. Using the data we estimated models of the speed vs IPC, and estimated the energy consumption of the vehicle for different trials. It was shown that using 10 trials at variable speed for each terrain the error between the measured energy and the estimated energy was 17%, furthermore, estimating the energy using the manufacturer speci cations gave an error up to 280%. We also estimated the energy in a longer path with 3 different terrains in it, getting an of 11.59%. We show that the models are an excellent tool to roughly estimate the energy consumption of a vehicle, despite being built mostly with constant speed data and all the trials for validation done at variable speeds. The method proposed in this work is easy to implement, to obtain the data and to generate the models. It is also possible to improve the models by including more variables in the measurements i.e. weight, slope angle, acceleration, etcetera. Compared to most of the works in the state of the art, that use IPC models with variables highly complex to calculate and estimate, our method is simpler and easily implementable in any kind of EV.