Machine Learning Approach for Modeling the Lateral Movement Decisions of Vehicles in Heterogeneous Traffic Conditions

The article describes the modeling of lateral movement decisions of motorized passenger vehicles like Cars, Motorised Three Wheelers(3W), and Motorised Two Wheelers(2W) under heterogeneous traffic conditions. The supervised machine learning approach was used to predict the lateral movement decision by treating the decision of lateral movement as a multi-class classification problem. Based on surrounding vehicles' information, a set of parameters was identified that potentially affect the decision-making process of drivers to change their lateral position. With the help of these parameters, the prediction ability of machine learning algorithms was compared. It was identified that these algorithms could predict the lateral movement decision of vehicles with an acceptable accuracy range. The results revealed that the random forest method out-performed all other algorithms and appeared to be a potential contender for modeling lateral movement decisions. Real-time position information about nearby vehicles may be gathered using advanced sensors and analyzed using developed models, allowing for the provision of safety features linked to lateral movement.