Qizhi Yu

Shanghai Zhangjiang Laboratory, Zhejiang Lab

Papers

Total Citations

147

H-Index

About

Qizhi Yu is a leading researcher in embodied AI and visual navigation, whose work bridges the gap between human-like spatial reasoning and autonomous robotic systems. His most influential contribution, the SOON framework (Scenario Oriented Object Navigation with Graph-based Exploration), tackles one of robotics’ most fundamental challenges: enabling agents to navigate toward language-guided targets from arbitrary starting points in complex 3D environments. With 115 citations, this work introduced a graph-based exploration strategy that dramatically improves generalization beyond fixed-startpoint benchmarks, setting a new standard for zero-shot navigation capabilities. Yu further advanced human-robot interaction through his depth-aware gaze-following research, which predicts where a person is looking in an image by integrating depth and orientation cues without requiring additional training datasets. This 2022 work, cited 32 times, streamlines the inference process for social robots that must interpret human attention in real time. His research consistently emphasizes practical, deployable solutions—from graph-based planning to auxiliary network architectures—that push toward the ‘holy grail’ of intelligent, context-aware robots capable of navigating and interacting in unstructured environments.