Online Data-driven Input Mapping Mixed Platoon Control with Variable Order Model

This paper investigates the longitudinal control problem of the mixed platoon consisting of human-driven vehicles (HDVs) and connected autonomous vehicles (CAVs) in the multi-lane freeway. In most of the existing studies, the stochastic behavior of HDVs, including lane changing and cut-in, are not considered and the formation of the mixed platoons remains unchanged. Also, the online identification of HDV uncertain parameters requires heavy computation. To address these limitations, this paper proposes an online data-driven control strategy for mixed platoon control with a variable-order model. This approach represents the uncertainties of HDVs in the form of convex hulls, and an event trigger mechanism is designed to deal with the stochastic behavior of HDVs that affects the formation and system model of mixed platoons. A data-driven control scheme synthesizing offline and online strategies is designed to deal with the uncertainties. The online part directly maps historical data to compensate for the offline control law without model identification, where the mapping coefficients can be efficiently calculated by a linear programming problem. The simulation results show that the proposed method is computationally more efficient and converges faster than the previous data-driven model predictive control and adaptive control methods.