Data-driven parameter estimation for optimal connected cruise control

In this paper we propose a data-based algorithm to identify parameters in a human car-following model, in order to facilitate the implementation of connected cruise control in real traffic. We first present a four-car experiment where the trajectory of each vehicle is recorded. Using the experimental data we identify the car-following parameters for each driver. Using the mean values of human parameters, we implement an optimal connected cruise controller on the last vehicle in the four-car string. We demonstrate by numerical simulation that the optimal connected vehicle design based on human parameter estimation has much smaller variations in headway and velocity, and acceleration compared with the human driver.