Publications

To boldly go where no man has gone before.

23 Citations 9 GitHub stars

2026

  1. arXiv
    Sim2Real-AD: A Modular Sim-to-Real Framework for Deploying VLM-Guided Reinforcement Learning in Real-World Autonomous Driving
    Sim2Real-AD: A Modular Sim-to-Real Framework for Deploying VLM-Guided Reinforcement Learning in Real-World Autonomous Driving
    Zilin Huang, Zhengyang Wan, Zihao Sheng, and

    Sim2Real-AD is a modular framework for transferring CARLA-trained, VLM-guided reinforcement-learning policies to full-scale vehicles without real-world RL training data. It supports zero-shot real-world deployment despite simulator-specific observation and action-semantics mismatches.

    arXiv preprint arXiv:2604.03497 · 2026
    Paper
    Scholar 0
  2. arXiv
    DriveVLM-RL: Neuroscience-Inspired Reinforcement Learning with Vision-Language Models for Safe and Deployable Autonomous Driving
    DriveVLM-RL: Neuroscience-Inspired Reinforcement Learning with Vision-Language Models for Safe and Deployable Autonomous Driving
    Zilin Huang, Zihao Sheng, Zhengyang Wan, and

    DriveVLM-RL combines a CLIP-based static safety pathway with attention-gated multi-frame VLM reasoning to learn semantic rewards for autonomous driving. Its asynchronous training pipeline removes all VLM components at deployment while improving collision avoidance, task success, and generalization in CARLA.

    arXiv preprint arXiv:2603.18315 · 2026
    Paper
    Scholar 3

2025

  1. JICV
    Sky-Drive: A Distributed Multi-Agent Simulation Platform for Socially-Aware and Human-AI Collaborative Future Transportation
    Sky-Drive: A Distributed Multi-Agent Simulation Platform for Socially-Aware and Human-AI Collaborative Future Transportation
    Zilin Huang*, Zihao Sheng*, Zhengyang Wan*, and

    Sky-Drive is a distributed multi-agent simulation platform for socially aware driving and human-AI collaboration. It combines synchronized multi-terminal simulation, multimodal human-in-the-loop data collection, adaptive human-AI knowledge exchange, and digital twins of real transportation environments.

    Journal of Intelligent and Connected Vehicles · 2025
    Paper
    Scholar 20 GitHub 9 ★

2024

  1. J. Automob. Eng.
    Study of the Tire Wear of Virtual Track Train
    Study of the Tire Wear of Virtual Track Train
    Hechao Zhou, Zhengyang Wan, Mingsu Mei, and

    This study develops and validates a multibody dynamics model of a Virtual Track Train together with a finite-element tire model to calculate tire wear. The results identify tire pressure, vehicle load, pavement conditions, inclination angle, and tire slippage as critical influences on wear.

    Proceedings of the Institution of Mechanical Engineers, Part D: Journal of Automobile Engineering · 2024
    Paper
    Scholar 0
  2. CSCD
    Development of Driving Simulation Platform for Virtual Track Train
    Development of Driving Simulation Platform for Virtual Track Train
    Zhengyang Wan, Hechao Zhou, and Jimin Zhang✉

    This work develops a Virtual Track Train driving simulator that combines a SIMPACK dynamics model, model-predictive control with Logitech driving hardware, and an Unreal Engine and Blender visualization interface. Simulations show that the platform captures operating performance across different driving modes.

    Urban Rapid Rail Transit · 2024
    Paper