I am a Ph.D. student in Trustworthy Autonomous Cyber-Physical Systems (ACPS) Lab at University of Liverpool with Prof. Xiaowei Huang and Dr. Xingyu Zhao. Prior to my Ph.D. program, I was working on vehicle transmission control supervised by Prof. Xiangyang Xu and Assoc. Prof. Peng Dong at Beihang University.

My research interests include Safe Reinforcement Learning, LLM-based Autonomous Driving, and Sequential Decision Making. I am eager to deploy Robust and Generalised algorithms to real robotics systems, like autonomous vehicles. The reference literature can be found in my Research Bibliography.

Contact:

Email: sihao.wu[at]liverpool.ac.uk

Office: Digital Innovation Facility, G018, Liverpool, UK

Profiles:

Github, GoogleScholar, Linkedin

Publications

Conferences

Robust RL with LLM-Driven Data Synthesis and Policy Adaptation for Autonomous Driving Sihao Wu Jiaxu Liu, Xiangyu Yin, Guangliang Cheng, Xingyu Zhao, Meng Fang, Xinping Yi, Xiaowei Huang Preprint, 2024 Paper This paper introduces RAPID to leverage the knowledge from LLMs to train an efficient and robust RL agent.

Data Augmentation for Continual RL via Adversarial Gradient Episodic Memory Sihao Wu, Xingyu Zhao, Xiaowei Huang 31st International Conference on Neural Information Processing (ICONIP2024), 2024 Adversarial Representation Training for Goal Conditioned Reinforcement Learning.

Tiny Refinements Elicit Resilience: Toward Efficient Prefix-Model Against LLM Red-Teaming Jiaxu Liu, Xiangyu Yin, Sihao Wu, Jianhong Wang, Meng Fang, Xinping Yi, Xiaowei Huang Paper Lightweight and General Protection layer against Red-Teaming attack for LLM.

A survey of safety and trustworthiness of large language models through the lens of verification and validation Xiaowei Huang, Wenjie Ruan, Wei Huang, Gaojie Jin, Yi Dong, Changshun Wu, Saddek Bensalem, Ronghui Mu, Yi Qi, Xingyu Zhao, Kaiwen Cai, Yanghao Zhang, Kaiwen Cai, Sihao Wu, Peipei Xu, Dengyu Wu, Andre Freitas, Mustafa A. Mustafa Artificial Intelligence Review, 57(7), 175. 2024 Paper Examining vulnerabilities and exploring how Verification and Validation techniques can be adapted to ensure LLM alignment with safety requirements.

Continuous Geometry-Aware Graph Diffusion via Hyperbolic Neural PDE Jiaxu Liu, Xinping Yi, Sihao Wu, Xiangyu Yin, Tianle Zhang, Xiaowei Huang, Jin Shi 2024 European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases (ECML PKDD), 2024 Paper, Github This paper reformulates Hyperbolic Graph Neural Networks as PDEs with node-wise attention, improving performance and scalability.

ReRoGCRL: Representation-based robustness in goal-conditioned reinforcement learning Xiangyu Yin, Sihao Wu, Jiaxu Liu, Meng Fang, Xingyu Zhao, Xiaowei Huang, Wenjie Ruan Proceedings of the AAAI Conference on Artificial Intelligence (AAAI), 2023 Paper, Github Adversarial Representation Training for Goal Conditioned Reinforcement Learning.

Design and Simulation of an Autonomous Racecar: Perception, SLAM, Planning and Control Sihao Wu, Zhengwei Yang, Xiaopo Xie, Yilong Wang, Xinliang Wang, Qi Wang, Bofan Wu, Hongjun Zhang, Hanning Zhang, Haochun Ma, Xuanliang Zhang, Haiying Lin IEEE International Conference on Autonomous Systems (ICAS), 2021 Paper The hardware and software concept of the 2020-rebuilt AERO Driverless Racing Team.

Journals

Coordinated clutch slip control for the engine start of vehicles with P2-hybrid automatic transmissions Peng Dong, Sihao Wu, Wei Guo, Xiangyang Xu, Shuhan Wang, Yahui Liu Mechanism and Machine Theory, vol.153, p.103899, 2020 (MMT) Paper Coordinated clutch control strategy for P2-hybrid transmissions.

Blogs!

Efficient and Generalized Deep Reinforcement Learning Sihao Wu Link Slightly change of environment will extremely impact RL agent performance, causing generalization problem.

Data Augmentation for Deep Reinforcement Learning Sihao Wu Link Data augmentation can be used to enforce the learning of an invariant representations.

Trustworthy ACPSLab introduction at DIF opening day Sihao Wu, Wei Huang, Yi Qi, Yi Dong Link Trustworthy ACPSLab is focusing on explaination, robustness, generalization, verification, validation and safety assurance for AI-based robotics.

Services