Manli Shu

Papers

Total Citations

H-Index

About

Manli Shu is a researcher whose work sits at the intersection of robust machine learning and 3D perception for autonomous systems. Her primary research areas include adversarial robustness, data augmentation, and 3D object detection from point clouds. Shu’s most cited work, “Adversarial Differentiable Data Augmentation for Autonomous Systems” (2021, 16 citations), tackles a critical vulnerability in neural networks used for planning and control: their severe performance degradation under input corruption not seen during training. This contribution is especially vital for real-world robotic systems operating in unpredictable environments. More recently, Shu has advanced 3D perception with her work on “Hierarchical Point Attention for Indoor 3D Object Detection” (2024, 6 citations), introducing a novel hierarchical structure to transformer architectures that improves detection in cluttered indoor spaces—a key capability for augmented reality and domestic robots. By addressing both the reliability and perceptual accuracy of autonomous systems, Shu is helping to bridge the gap between controlled lab settings and the messy, dynamic conditions of the real world.