Nonlinear Data-Driven Predictive Control for Mixed Platoons Based on Koopman Operator

Mixed traffic, where Intelligent and Connected Vehicles (ICVs) and Human-Driven Vehicles (HDVs) coexist, is expected to persist considerably. In such scenarios, mixed platoon control strategies have shown promise for improving traffic performance. However, most existing studies have focused on control methods based on linear systems, even though real-world mixed platoon systems exhibit significant nonlinear characteristics. To tackle this challenge, this paper introduces a Koopman Model Predictive Control (MPC) approach to ensure safe and optimal control for mixed platoon systems. Specifically, this method utilizes Koopman theory to transform the original system states into a high-dimensional space, deriving an approximate data-driven Koopman-based predictor of the original nonlinear system. This predictor is then integrated into the MPC framework, formulating a Koopman MPC optimization problem that considers system safety constraints explicitly. Nonlinear simulation tests demonstrate that the proposed Koopman MPC method significantly outperforms baseline methods in improving the tracking performance of mixed platoons.