Biography

Hi! I am Liu Zihao, currently a PhD student in control science and engineering at the School of Automation, Northwestern Polytechnical University, with Huang Panfeng and Liu Xing as my supervisors.

I have won championships in the China Robot Competition and the China Robot and Artificial Intelligence Competition, and third prize in the Robotmaster International Competition. I have also won the first-class scholarship, Kewei Scholarship, and Wu Yajun Scholarship at Northwestern Polytechnical University many times.

My research interests are mainly focused on robot manipulation, especially embodied AI, imitation learning, reinforcement learning and tactile sense.

Publications

Curiosity-Diffuser: Curiosity Guide Diffusion Models for Reliability

Zihao Liu, Xing Liu, Yizhai Zhang, Zhengxiong Liu, Panfeng Huang

In order to avoid hallucinations in the output of the policy, the diffuser model is improved to guide the output of the policy close to its training set.

STAGE: STyle driving Action GEneration with preference imitation learning

Zihao Liu, Xing Liu, Yizhai Zhang, Panfeng Huang

Combining multimodal imitation learning and preference learning, a driving style conditioned driving policy network is created to achieve efficient human-machine collaboration.

Tactile Active Inference Reinforcement Learning for Efficient Robotic Manipulation Skill Acquisition

Zihao Liu, Xing Liu, Yizhai Zhang, Zhengxiong Liu, Panfeng Huang

IROS 2024

By combining active inference reinforcement learning with the sense of tactile in the context of the manipulation, we obtain a very data-efficient method for learning manipulation tasks by reinforcement learning.

Efficient Reinforcement Learning Method for Multi-Phase Robot Manipulation Skill Acquisition via Human Knowledge, Model-Based, and Model-Free Methods

Xing Liu, Zihao Liu, Gaozhao Wang, Zhengxiong Liu, Panfeng Huang

IEEE trans TASE

A combination of human knowledge, model-based, and model-free approaches is used to achieve robotic manipulation control in complex multi-stage robotic manipulation tasks such as peg-in-hole assembly, screwing nuts, etc. Participated in the development of the PILCO algorithm and reinforcement learning screwing task.

Club

2020.9~2023.6 Soccer Robot Club V5++ Group

2019.9~2020.6 ACM Algorithm Training Team