Mingjie Sun

Papers

Total Citations

H-Index

About

Mingjie Sun is a rising researcher whose work spans the critical intersection of artificial intelligence security and 3D perception. His early, highly-cited research on "Characterizing Attacks on Deep Reinforcement Learning" (2019, 52 citations) established foundational insights into the vulnerabilities of DRL models, systematically analyzing how adversarial perturbations can compromise decision-making in autonomous systems. This work highlighted the practical limitations of existing attack methods—such as their reliance on full model access and prohibitive computational costs—paving the way for more realistic threat models. More recently, Sun has ventured into robust 3D scene understanding, as evidenced by his 2025 paper on a "Simple MLP Framework for Z-Axis Rotation-Invariant Point Cloud Place Recognition." This work addresses a persistent challenge in robotics and autonomous navigation: achieving reliable place recognition despite rotational variations in LiDAR scans. By demonstrating that a streamlined MLP architecture can outperform complex deep learning models on this task, Sun is contributing to more efficient and generalizable perception systems. His trajectory from adversarial robustness to geometric invariance signals a commitment to building AI that is both secure and spatially intelligent.