Classification of adaptive cruise control vehicle type based on car following trajectories

The emergence of vehicles with driver-assist features or other automated driving capabilities has been shown to dramatically alter traffic dynamics at the aggregate level. These new traffic flow dynamics usher in the possibility for specialized traffic control based on the types of vehicles with differing car following dynamics and the resulting shift and macroscopic traffic flow behavior. To enable such control, vehicle identification techniques are needed that allow realtime identification of the types of vehicles in the traffic flow to predict the aggregate behavior of the resulting flow. This work presents a novel, model-based approach to classify the type of adaptive cruise control (ACC) equipped vehicle based on the vehicle car following dynamics. The approach first calibrates best-fit car following model parameter values based on observed car following trajectories and classifies the vehicle type based on the calibrated model parameter values. Both the k nearest neighbor and support vector machine classifiers are tested on an experimental vehicle trajectory dataset. The support vector machine classifier is found to perform best. This result enables rapid, real-time classification of ACC vehicle type which will enable the next generation of traffic control based on the individual constituents of the traffic flow.