Li, Lei (Southeast University), Xu, Liwei (Southeast University), Liu, Zilong (Southeast University), wang, tingyuan (Southeast University), Li, Ang (Southeast University), Yin, Guodong (Southeast University)

Deep-Koopman-Based Model Predictive Control for Trajectory Tracking of Automated Vehicle

Scheduled for presentation during the Invited Session "S21c-Energy-Efficient Connected Mobility" (TH-LA-T21), Thursday, November 20, 2025, 16:40−17:00, Surfers Paradise 3

2025 IEEE 28th International Conference on Intelligent Transportation Systems (ITSC), November 18-21, 2025, Gold Coast, Australia

This information is tentative and subject to change. Compiled on October 18, 2025

Abstract

This paper proposes a novel trajectory tracking model predictive control (MPC) method leveraging deep neural network-based Koopman operator theory. Specifically, the approach employs a deep-learning-enhanced Extended Dynamic Mode Decomposition (Deep EDMD) framework, integrating encoder-decoder architectures to automatically learn nonlinear lifting and projection functions for vehicle dynamics. The resulting linear representation significantly simplifies control computations. Extensive simulations in a high-fidelity CarSim/Matlab environment demonstrate robust tracking performance under dynamic curvature scenarios, with maximum lateral displacement errors remaining within ISO 3888-2 standards. Comparative analyses reveal that the proposed method substantially outperforms traditional MPC, exhibiting improved stability, tracking accuracy, and adaptability under varying adhesion coefficients and vehicle speeds.