Zetao Du

Papers

Total Citations

H-Index

About

Zetao Du is a rising researcher in embodied AI and multimodal learning, with a focus on bridging perception and action in real-world environments. His most-cited work, "Multimodal Pretrained Knowledge for Real-world Object Navigation" (2025), introduces a novel framework that leverages pretrained vision-language models to guide robotic agents in locating objects in dynamic, unstructured settings. This contribution addresses a critical bottleneck in autonomous navigation—how to transfer knowledge from static datasets to interactive, physical spaces—by integrating semantic reasoning with spatial planning. While early in his career, Du's work has already garnered attention for its practical implications in service robotics and assistive technology. His research sits at the intersection of computer vision, natural language processing, and robotics, aiming to create agents that understand and act upon human-like cues. As a young scholar, Du’s approach exemplifies how multimodal pretraining can be repurposed beyond recognition tasks, offering a scalable path toward more adaptable and intelligent robots. His ongoing work promises to further bridge the gap between simulated training and real-world deployment.