Physics-Informed Neural Networks for Learning the Parameters of Commercial Adaptive Cruise Control Systems

This paper develops a physics-informed neural network (PINN) for learning the parameters of commercially implemented adaptive cruise control (ACC) systems. The constant time-headway policy (CTHP) is adopted to emulate the core functionality of stock ACC systems (proprietary control logic and its parameters) which is not publicly available. Multi-layer artificial neural networks is a class of universal approximators, and thus the developed PINN can serve as a surrogate approximator to capture the longitudinal dynamics of ACC-engaged vehicles and efficiently learn the unknown parameters of the CTHP. The ability of the PINN to infer the unknown ACC parameters is tested on both synthetic and empirical data of space-gap and relative velocity involved ACC-engaged vehicles in platoon formation. The results have demonstrated the superior predictive ability of the proposed PINN to learn the unknown design parameters of stock ACC systems of different vehicle makes. The set of ACC model parameters obtained from the PINN revealed that the stock ACC system of the considered vehicles in three experimental campaigns is neither $\mathcal{L}_{2}$ nor $\mathcal{L}_{\infty}$ string stable.