Online parameter estimation of adaptive cruise control models with delays and lags

In this work we consider online parameter estimation of adaptive cruise control (ACC) equipped vehicles which may contain time delays (e.g., sensor delays) and lags (e.g., actuator lag). We extend a recursive least squares (RLS) method and apply it to calibrate models with various magnitudes of delay and lag, and study the performance of the method when measurement noise is present. We show that RLS can be applied to exactly recover model parameters given noise-free measurement data. In addition, the RLS estimator performs well under typical sensor noises associated with onboard sensors such as GPS and radars. The method is tested on data from a 2019 ACC-equipped vehicle, and the results show that the overall quality of fit via RLS is comparable to a commonly used batch optimization method (4% and 0.8% mean absolute error on the space gap and the velocity profiles, respectively). RLS is shown to run two orders of magnitude faster than the batch optimization method.