Drag coefficient modeling of heterogeneous connected platooning vehicles via BP neural network and PSO algorithm

The operation of connected platooning vehicles (CPVs) with V2V and V2I techniques has the potential to decrease the overall aerodynamic drag and reduce fuel consumption. One of the key technologies for CPVs is the spacing control which is heavily dependent on the drag force. However, the existing studies on drag coefficient of CPVs seldom pay attention to heterogeneous platoon, and most analysis only consider the effect of vehicle spacing and ignore key information such as cross-sectional area of adjacent vehicles. In addition, most research only focuses on qualitative research, without concluding a quick formula for calculating vehicle drag coefficient in a platoon. In this work, we investigate the modeling and estimation of drag coefficient in terms of the inter-vehicle distances, the platoon configurations, and the cross-sectional area of the adjacent vehicles. We calculate the drag coefficients for ten categories of vehicles with different inter-vehicle distances via numerical analysis in the software FLUENT. According to hundreds of the analysis results, we propose a model to estimate the drag correction factor online. A hybrid algorithm combining the BP neural network (BPNN) and particle swarm optimization (PSO) is employed to optimize the parameters in the model. The model is validated via three kinds of regression evaluation indexes and additional simulation tests with different platooning configurations and different types of vehicles. In terms of the comparison results, the developed model is effective to estimate the drag coefficients of CPVs.