IROS 2025 Keynotes - Embodied Intellgence: Fumiya Lida

Why It Matters

By marrying dense physical embodiment with AI principles, robotics can achieve the adaptability and robustness needed for real‑world deployment, narrowing the gap between artificial and biological intelligence.

Key Takeaways

• Embodied intelligence links body mechanics with brain-like control loops.
• Human bodies contain 30 trillion cells, far surpassing robot complexity.
• Simple sensor‑free robots can exhibit complex behaviors via embodiment.
• Multi‑material fabrication enables passive, adaptive physical intelligence in robots.
• Future research aims to create embodied transformers mimicking reflexive attention.

Summary

Fumiya Lida’s IROS 2025 keynote framed embodied intelligence as the reciprocal relationship between a body’s physical dynamics and the brain’s control mechanisms, challenging the longstanding brain‑versus‑body dualism that has split robotics from AI. He highlighted the staggering scale gap—30 trillion cells in a human body versus a few million parts in the largest robots—underscoring how much richer biological embodiment is compared to current machines. Lida illustrated his points with a series of robots ranging from “brainless” sinusoidal locomotors that walk, hop and dance without sensors, to continuum‑beam walkers and multi‑material structures that exploit passive dynamics. A striking example was a soft hand equipped with 32 barometric sensors, allowing it to anticipate grasp failures before they occur, demonstrating how dense sensing and physical compliance can replace complex computation. He drew a provocative parallel between transformer models in generative AI and physical attention mechanisms in bodies, proposing an “embodied transformer” that would use reflex‑like pathways to process sensory input and generate motor output. This concept builds on the idea that reflex circuits achieve rapid, context‑aware responses without heavy learning, offering a blueprint for next‑generation robots. The talk concluded that to compete with the impact of generative AI, robotics must broaden its focus from narrow algorithmic tricks to holistic physical intelligence—leveraging soft materials, dense sensing, and embodied attention. Such integration promises more adaptable, resilient machines capable of operating in the messy real world, potentially reshaping manufacturing, service robotics, and human‑machine interaction.

Original Description

"Keynote Title: ""Informatizing Soft Robots for Super Embodied Intelligence""

Speaker Biography

Fumiya Iida is a Professor at School of Engineering, the University of Tokyo, a director of Research at University of Cambridge, and the director of Bio-Inspired Robotics Laboratory. He received his bachelor and master degrees in mechanical engineering at Tokyo University of Science (Japan, 1999), and Dr. sc. nat. in Informatics at University of Zurich (2006). In 2004 and 2005, he was also engaged in biomechanics research of human locomotion at Locomotion Laboratory, University of Jena (Germany). From 2006 to 2009, he worked as a postdoctoral associate at the Computer Science and Artificial Intelligence Laboratory, Massachusetts Institute of Technology in USA. In 2006, he awarded the Fellowship for Prospective Researchers from the Swiss National Science Foundation, and in 2009, the Swiss National Science Foundation Professorship for an assistant professorship at ETH Zurich until 2015. He was also a Professor of Robotics at the University of Cambridge until 2025. He was a recipient of the IROS2016 Fukuda Young Professional Award, Royal Society Translation Award in 2017, Tokyo University of Science Award in 2021. His research interest includes biologically inspired robotics, embodied artificial intelligence, and biomechanics, where he was involved in a number of research projects related to dynamic legged locomotion, dextrous and adaptive manipulation, human-machine interactions, and evolutionary robotics.

Abstract

Soft robotics has made remarkable advances in developing deformable functional materials for locomotion, manipulation, and other forms of morphological adaptation such as self-morphing, self-healing, and mechanical growth. While these technologies have opened up new applications for robotics, they also present novel challenges in sensing, modelling, planning, and control. Due to the inherent complexity of systems based on flexible and continuum mechanics — and the wide range of interactions with their environments — conventional methods often fall short, making novel approaches rooted in advanced machine learning essential. In this talk, I will introduce several projects in our laboratory that leverage sensorized soft robots and machine learning to tackle these challenges. I will also present the concept of “Super Embodied Intelligence” as a new research framework for realizing the next generation of intelligent robots and its technological underpinnings. As research in soft robotics and functional materials progresses, we are witnessing a fusion of the informational and physical entities. Within this context, where new forms of embodied intelligence are emerging, I will discuss how rapidly evolving fields such as machine learning can accelerate this development. Moving beyond traditional notions of bodily control and AI as purely computational, this approach explores the potential for new forms of intelligence in which the body itself becomes an active site for information processing and generation.

"