SoftMechs: Mechanically Intelligent Soft Robots
Abstract: Inspired by nature, soft robots built primarily from compliant materials aim to emulate the remarkable capabilities of biological systems, especially in unstructured and complex environments. Yet, they fall short in matching the structural adaptability, physical robustness, and power autonomy (the ability to operate indefinitely without manual energy input) of their natural counterparts. My research aims to bridge this gap via Mechanical Intelligence (MI)—a strategy that embeds sensing, control, actuation, and power directly within robots' mechanical structures, thereby reducing dependence on computational intelligence (CI), like high-performance processors and AI. By adopting MI, we can simplify, minimize, and strengthen current soft robots that rely heavily on CI, which can be inefficient or impractical in complex environments. The resulting mechanically intelligent soft robots, or SoftMechs, will leverage AI in design and selectively integrate AI for optimized functionality, ultimately moving toward truly robust, adaptive, and autonomous agents that mirror the capabilities of biological organisms.
In this talk, I will demonstrate the power of MI, focusing on three thrusts: the power, body, and control. First, I will present methods for harvesting constant environmental energy via functional materials and structures into useful formats for independently powering robots. I’ll demonstrate this by building an autonomous soft robot that can continuously jump using energy harvested from light, without requiring a tether or external control. Then, I will showcase how to improve the structural adaptability of soft robots so they can adapt to changing environments while maintaining loading capability. I’ll introduce an adaptive metamaterial capable of adjusting its shape and stiffness as needed. Building on this, I will show you a soft robot that can crawl through restricted, complex spaces, adapting its body to the environment. I will also demonstrate how to embed autonomous behaviors directly into mechanical structures, making them robust in adversarial environments, especially where semiconductors would fail. I will showcase this using soft mechanical circuits and several purely mechanical robots. One example, inspired by the Venus flytrap, can recognize and capture "living" objects without the need for semiconductor-based electronics. Finally, I will conclude with a discussion on integrating advancements in power, body, and control, enabling the creation of truly self-sustained, adaptive, and robust soft robots that not only mirror but may even exceed the capabilities of biological organisms
Bio: Wenzhong Yan is a postdoctoral fellow in the Electrical and Computer Engineering Department at UCLA, working with Ankur Mehta. He earned his Ph.D. in Mechanical and Aerospace Engineering from the same institution in 2023. His research interests are broad, encompassing soft robotics, origami robots, mechanical intelligence, energy harvesting, structural instability, computational fabrication, multifunctional materials and actuators, and wearable devices. Wenzhong has received several prestigious awards, including the Cyber-Physical Systems Rising Star, Robotics: Science and Systems Pioneer, and the 2024 Edward K. Rice Outstanding Doctoral Student Award from UCLA's Samueli School of Engineering. His work has been published in top journals such as Science, Nature Communications, Science Advances, Materials Horizons, and Soft Robotics, and has been featured in prominent press outlets like National Science Foundation News, EurekAlert (AAAS), and Popular Science.
Zoom link: https://wpi.zoom.us/j/93413349160