Research and Academia Agriculture Robots

AI-Embodied Flexible Electronic Robots Break New Ground in Miniature Autonomy

Explore how AI-embodied flexible electronic robots (FEBots) fuse sensing, actuation, and self-learning into tiny, autonomous soft machines that navigate complex environments with insect-like agility, pointing to future applications in search-and-rescue, inspection, and medical micro-robotics.

Share
AI-Embodied Flexible Electronic Robots Break New Ground in Miniature Autonomy
Share

From insect agility to intelligent autonomy

One of the grand frontiers in robotics is to build microscale robots that match the agility and autonomy of insects — capable of navigating cluttered, unstructured environments that defy conventional engineering.
Whether exploring collapsed buildings for survivors or inspecting narrow industrial pipelines, such robots must seamlessly combine terrain adaptability, real-time sensing, and on-board decision-making.
For soft robots, however, integrating all three functions — locomotion, perception, and computation — within a tiny and lightweight body has long remained a formidable challenge.

A research team from Huazhong University of Science and Technology (HUST) and collaborators has now bridged that gap by fusing flexible electronics with oscillatory actuation into a unified intelligent platform.
Their achievement, published in Nature Communications under the title AI-embodied multi-modal flexible electronic robots with programmable sensing, actuating and self-learning, represents a significant step toward embodied AI in soft robotics.  

Rearch Team: Junfeng Li, Zhangyu Xu, Nanpei Li, Kaijun Zhang, Guangyong Xiong, Minjie Sun, Chao Hou, Jingjing Ji, Fan Zhang, Junwen Zhong & YongAn Huang 

 

The FEBot: Modular design meets flexible intelligence

At the heart of this breakthrough is the Flexible Electronic Robot (FEBot) — a modular, reconfigurable system that merges structural adaptability with embedded computation.

Unlike legged robots that rely on precise gait control, FEBots use distributed setae arrays — micro bristle structures inspired by insect feet — to achieve motion through asymmetric friction rather than complex joint coordination.
This minimalist approach simplifies control while dramatically improving adaptability to uneven terrain.

Each FEBot consists of two main parts:

  • Programmable Flexible Electronic Modules, including:
    1. Strain-sensing actuators
    2. Temperature and humidity sensors
    3. Proximity sensors
    4. A central controller integrating an NRF52832 microchip and MPU6050 IMU
      These modules are linked via conductive adhesive pads, enabling true plug-and-play functionality and rapid reconfiguration.
  • Distributed Setae Arrays, made of superelastic shape memory alloy (SSMA), offering excellent elasticity, corrosion resistance, and durability.
    These bristles are not passive — they serve as the core of the robot’s actuation mechanism.

By combining these modules, researchers can rapidly prototype robots with different morphologies — such as a millipede-like Type I design optimized for confined spaces, and a square Type II model designed for stable outdoor navigation.

 

Oscillatory actuation and the physics of asymmetric friction

The FEBot’s locomotion stems from a biomimetic oscillation mechanism driven by periodic deformation and recovery of its SSMA bristles.
The key lies in directional friction asymmetry: friction differs when the robot slides forward versus backward.

Through both modeling and high-speed imaging (3,000 fps), the team mapped a complete actuation cycle:

  • Compression phase (State I) – Bristles are fully pressed down.
  • Backward slip (State II) – As the unit rises, backward friction briefly dominates, producing a tiny rearward displacement (~0.085 mm).
  • Forward slip (State III) – During descent, lower forward friction allows a larger forward step (~0.265 mm).

This imbalance yields a net forward motion per cycle.
Using Cosserat rod theory and a spring-damper model, simulations matched experimental trajectories with remarkable precision.

Parameter sweeps revealed that bristle geometry critically determines speed and stability.
An optimal configuration — length L = 7 mm, diameter d = 0.1 mm, contact angle θ = 60° — delivered a peak velocity of 109.5 mm/s and a maximum climbing angle of 18°, achieving the best balance among speed, stability, and terrain adaptability.

 

Multi-modal motion and environmental awareness

What truly sets FEBots apart is their multi-modal mobility and environmental intelligence.
Their modular architecture enables Lego-style reconfiguration for mission-specific behavior.

The Type I millipede-inspired robot, for instance, can crawl at 87.6 mm/s through narrow vertical channels, carrying loads up to 5.1 times its body weight — and squeezing through gaps only 70 % of its body width (14 mm).
Meanwhile, the Type II square configuration supports omnidirectional motion, capable of turning or spinning in place.
Encapsulated with a waterproof coating, it even traverses underwater surfaces at 9 mm/s, showcasing cross-medium locomotion.

An ingenious foldable bristle mechanism — driven by thermal actuation of shape-memory alloy springs — allows the bristle angle to shift from 0° to 45°, reversing movement direction on command.
This mechanical design endures pressure up to 250,000 times the robot’s weight without damage, demonstrating exceptional resilience.

Beyond locomotion, FEBots are equipped with a multi-sensor suite — inertial, strain, temperature, humidity, and proximity sensors, and even miniature cameras — enabling simultaneous monitoring of internal state and external environment.
Coupled with embedded computation and adaptive control algorithms, these robots exhibit primitive self-learning behavior, adjusting their motion based on sensor feedback.

Toward embodied AI at the microscale

The FEBot platform illustrates how embodied intelligence — computation physically interwoven with material and structure — can empower soft robots to function autonomously in the wild.
By uniting programmable sensing, adaptive actuation, and data-driven self-learning in a single flexible form, HUST’s team brings robotic autonomy closer to the insect scale — where intelligence is not only coded, but also materialized.

This research opens pathways for next-generation field robots, from medical microrobots to disaster-response swarms, where miniature intelligence meets material adaptability.

image.png

RobotToday Initiative

Robotics needs a service framework.

RSF defines a common language for robot service capability, lifecycle operations, certification pathways, and service-provider networks.

Share
Written by
Kelly Stone - Associtae Editor

Kelly Stone is an Associate Editor focused on industrial technology, covering robotics, automation systems, and AI applications. Her reporting emphasizes company funding, market structure, and emerging industry trends. She has three years of experience in technology media.