Yoshihiko Hayashi

Papers

Total Citations

H-Index

About

Yoshihiko Hayashi is a leading researcher in human-robot interaction and grounded language learning, with a focus on bridging the gap between linguistic symbols and physical robot actions. His most influential work, "Embodying Pre-Trained Word Embeddings Through Robot Actions" (2021, 19 citations), introduces a novel neural network model that enables robots to ground abstract language—including polysemous words—in real-world actions, a critical step toward intuitive human-robot collaboration. Building on this, his 2022 study (7 citations) achieves bidirectional translation between natural language descriptions and robot actions using surprisingly small paired datasets, overcoming a major data scarcity bottleneck in multimodal learning. Hayashi also contributes to safety in shared spaces through his analysis of human avoidance behavior in narrow passages (2018, 2 citations), laying groundwork for robots that can predict and adapt to human movement. His work is notable for tackling the "symbol grounding problem" in robotics, demonstrating how pre-trained language models can be effectively embodied. With applications ranging from assistive robots to autonomous navigation, Hayashi’s research is shaping how machines understand and act upon human language in the physical world.