Haofeng Liu

South China Agricultural University, National University of Singapore

Papers

Total Citations

158

H-Index

About

Haofeng Liu is a researcher at the forefront of agricultural robotics and intelligent manipulation systems. His work bridges computer vision and robotic control, with a primary focus on enabling machines to perceive and interact with complex, unstructured environments. Liu’s most influential contribution is his 2020 study on "Passion fruit detection and counting based on multiple scale faster R-CNN using RGB-D images," which has garnered 154 citations. This work established a robust, multi-scale deep learning framework for automated fruit detection, addressing critical challenges in precision agriculture by combining RGB and depth data to improve accuracy in natural settings. More recently, Liu has advanced the field of robotic manipulation, tackling the persistent challenge of learning effective policies from limited demonstrations. His proposed methods, including E-GAIL (Efficient Generative Adversarial Imitation Learning) and deterministic policy approaches, introduce innovative techniques such as negative corruption and long-term reward structures to dramatically improve sample efficiency. These contributions are particularly significant for real-world applications where collecting large, high-quality demonstration datasets is impractical. Liu’s research represents a meaningful step toward more adaptable and data-efficient robotic systems, with implications spanning from automated harvesting to complex industrial manipulation tasks.