Papers
2
Total Citations
5
H-Index
2
About
Minjae Song is a rising researcher in embodied AI and human-robot interaction, with a focus on enabling robots to learn from human demonstrations and natural language. Their work bridges the gap between high-level human intent and low-level robotic control, particularly through inverse constraint learning and multimodal grounding. In their 2023 paper "Inverse Constraint Learning and Generalization by Transferable Reward Decomposition" (3 citations), Song introduced a novel framework to recover hidden constraints from demonstrations—addressing the ill-posed nature of inverse reinforcement learning—allowing robots to safely and autonomously reproduce skilled behaviors in unseen environments. This work is foundational for safe robot learning in dynamic settings. Additionally, in "SGGNet²: Speech-Scene Graph Grounding Network for Speech-guided Navigation" (2 citations), they tackled the challenge of grounding spoken language in visual scenes for assistive robotics, handling acoustic variability to make navigation accessible for non-experts and disabled users. Though early in their career, Song’s contributions are shaping more intuitive and robust human-robot collaboration, with clear potential for high impact in service robotics and autonomous systems.
Research Focus
Key Achievements
Top Papers
- 1
- 2