Vehicle Lane Change Decision Model Based on Random Forest

Traditional vehicle lane change decision model based on the mathematical model method and the logical model method is difficult to accurately describe the complete lane changing process and can't reflect a series of psychological and physiological reactions of the driver. And the subjective experience of the formulation model leads to the lack of certain rationality of this kind of models. This paper proposes a lane change decision model based on ensemble learning from the perspective of data driving. Based on the analysis of the decision-making factors of discretionary lane change of manual driving vehicles and the behavior characteristics of vehicles in the course of lane-changing execution, the identification method of key point in the process of autonomous lane change is proposed. The vehicle lane change behavior decision model based on random forest is established by selecting appropriate variables as decision factors. Then, the vehicle data of a driving unit is extracted from the NGSIM dataset. In order to correct the errors in the NGSIM data to obtain more accurate vehicle movement data, symmetric exponential moving average algorithm is used to smooth the extracted sample data. Finally, the random forest lane-changing decision-making model is trained and tested with the pre-processed sample data, and good prediction results and fitting degree are obtained.