Attribute selection for modelling driver's car-following behaviour in heterogeneous congested traffic conditions

This paper uses a real-world data set to investigate a driver's car-following behaviour of different class of vehicles in congested traffic conditions. The existing car-following models do not explicitly consider heavy vehicle (HV) interactions with the other vehicles. This could become problematic in future due to the increasing proportion of HVs in the traffic stream. Four types of vehicle combinations were considered in this study including car-car, car-HV, HV-car, and HV-HV. The results of detailed data analysis showed that the driver's behaviours differ in each car-following combination. Further the variables which could influence the car-following behaviour in each combination were identified. The potential variables were explored and the effective variables were selected through a combination of advanced statistical analysis. The findings specify that further research is needed to develop a car-following model which incorporates these behavioural differences.