کنفرانس مهندسی برق ایران

صفحه اصلی / سی و چهارمین کنفرانس بین المللی مهندسی برق

Dual-Network Learning-Enhanced DMPC for Mixed Autonomous–Human Vehicle Platoons

نویسندگان :

Parsa Akbari¹ Abolfazl Saadati Moghadam² Mohammad Haeri³

1- Sharif university of technology 2- Sharif university of technology 3- Sharif university of technology

کلمات کلیدی :

Mixed platoon،Learning-based control،Connected automated vehicles،Distributed model predictive control

چکیده :

Distributed model predictive control is an effective approach for connected autonomous vehicles platoons, but its performance degrades in mixed traffic due to the stochastic behavior of human-driven vehicles and heterogeneous vehicle parameters. This paper proposes a dual-network learning-enhanced distributed model predictive control framework for mixed autonomous–human platoons. A long short-term memory network is integrated to predict the residual behavior of human-driven vehicles, enabling proactive anticipation of human driving actions. In parallel, a multilayer perceptron replaces nonlinear vehicle dynamics constraints, yielding a model-free controller adaptable to parameter variations without explicit identification. A dynamic safety-switching strategy is further embedded in the distributed model predictive control to prioritize collision avoidance under hazardous conditions. Simulations in a mid-platoon scenario demonstrate improved transient response, enhanced safety, and reliable performance during aggressive human-driven vehicle braking events.