Fangru Zhou

Papers

Total Citations

H-Index

About

Fangru Zhou’s research lies at the intersection of robotics, computer vision, and embedded systems, with a primary focus on visual SLAM (Simultaneous Localization and Mapping) and real-time scene understanding. His most impactful contribution is a novel approach to visual loop closure detection, a critical problem in SLAM that ensures a robot can recognize previously visited locations. By leveraging proximity graphs, Zhou developed a fast and incremental method that significantly improves upon traditional bag-of-words models, achieving high precision with reduced computational overhead—work that has garnered over 43 citations. Building on this, Zhou addressed the pressing need for safe human-robot interaction with a low-complexity neural network architecture for monocular human depth estimation and segmentation. Designed specifically for embedded systems, this work enables real-time collision avoidance against moving pedestrians, a vital capability for autonomous robots operating in indoor environments. By balancing accuracy with the stringent constraints of onboard processing, Zhou’s research directly advances the practicality of autonomous navigation in human-centric spaces. His contributions are shaping the next generation of efficient, perception-driven robotics.