Junyi Ma

Papers

Total Citations

H-Index

About

Junyi Ma is a leading researcher in autonomous navigation and robotic perception, with a primary focus on place recognition—a critical capability for self-driving vehicles and robots operating in GPS-denied environments. Ma’s most impactful contribution is the **OverlapTransformer** (2022, 9 citations), a novel, rotation-invariant transformer network that dramatically improves LiDAR-based place recognition for tasks like loop closure in SLAM and global localization. This work set a new standard for efficiency and robustness under changing environmental conditions. More recently, Ma has advanced **fusion-based place recognition** (2024, 3 citations), pioneering explicit interaction techniques that intelligently combine multi-modal perception data—such as vision and LiDAR—to achieve more reliable localization than implicit fusion methods. Expanding into human-robot interaction, Ma’s **Diff-IP2D** (2024, 2 citations) applies diffusion models to predict hand-object interactions from egocentric video, forecasting both hand trajectories and object affordances for applications in service robotics and extended reality. With a growing citation impact and a clear trajectory from foundational perception to interactive prediction, Junyi Ma is shaping the future of how autonomous systems understand and navigate complex, dynamic environments.