⚡ electrical engineering

Hollow-strut lattice architectures for distributed actuation and sensing in monolithic soft robots

This paper presents a monolithic soft robot fabrication method using hollow-strut lattice architectures to integrate distributed pneumatic actuation and multi-mode sensing within a single printed body, thereby eliminating the need for post-assembly fluidic tubing while enabling simultaneous bending and tactile detection through a topology-aware, single-channel sensing system.

Original authors: Altair Coutinho, Trevor Exley, Petr Trunin, Anderson Brazil Nardin, Lucia Beccai

Published 2026-07-08

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Altair Coutinho, Trevor Exley, Petr Trunin, Anderson Brazil Nardin, Lucia Beccai

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine building a soft robot out of a single, continuous piece of rubber, like a giant, flexible Lego brick. Usually, when you make these robots, you have a problem: you need to pump air into different parts to make them move (actuation) and you need to squeeze them to make them "feel" things (sensing). But getting air tubes and sensors to all the right spots usually means gluing extra hoses and wires onto the robot after you print it. It's messy, fragile, and requires a lot of assembly.

This paper introduces a clever new way to build these robots so that everything is built-in from the start, with no extra tubes or wires needed.

Here is how they did it, using some simple analogies:

1. The Robot is a "Honeycomb" with Secret Tunnels

Instead of a solid block of rubber, the robot is built like a honeycomb (a lattice structure). The researchers made the "walls" of this honeycomb hollow, like tiny straws.

The Magic: These hollow straws aren't just for structure; they act as the robot's internal plumbing.
The Result: You don't need to glue tubes on the outside. The robot prints itself with a complete network of tunnels inside, ready to carry air.

2. Two Roads in One Tunnel System

The team figured out how to send two different types of traffic through this single honeycomb network without them crashing into each other:

The "Muscle" Road: One set of tunnels carries air to the robot's muscles (actuators) to make it wiggle and crawl.
The "Nerve" Road: A second, separate set of tunnels runs through the robot to act as a sensor. When the robot bends or gets squeezed, the air inside this tunnel gets squished, changing the pressure.

3. The "Volume Knob" Trick (Diameter Programming)

This is the most creative part. The researchers realized they could change how sensitive the "Nerve" road is just by changing the width of the tunnel in different spots.

Wide Tunnel (2.5 mm): Think of this as a thin-walled, sensitive balloon. If you squeeze it or bend the robot, the air inside squishes easily, creating a big pressure change. This is used for "feeling" things.
Narrow Tunnel (2.0 mm): Think of this as a thick-walled, stiff pipe. If you bend the robot, this part barely changes. It acts like a quiet hallway, just moving air from one place to another without "feeling" anything.

By printing the robot with a mix of wide and narrow sections, they can tell the robot exactly where to feel and where to just pass air along, all by changing the diameter of the hollow struts.

4. The "Two-Ear" Listening System

Since the robot uses air (which is squishy) instead of water, the pressure signals can be small. To solve this, they connected two different pressure gauges to the ends of the single sensing tunnel:

The Sensitive Ear: One gauge is very delicate. It hears the soft, gentle sounds of the robot bending.
The Loud Ear: The other gauge is built to handle loud, sudden noises.
How it works: When the robot gently bends, the "Sensitive Ear" hears it. But if you pinch the robot hard, the signal gets too loud for the sensitive gauge (it "saturates" or stops working). However, the "Loud Ear" catches that big pinch perfectly.
The Benefit: The robot can tell the difference between "I am bending" and "I am being pinched" just by looking at how loud the signal is, using only one single tunnel.

5. The "Crawling Robot" Demo

They built a small, inchworm-like robot to prove this works.

Movement: It crawls forward by pumping air into its "Muscle Road."
Sensing: As it crawls, the "Nerve Road" tells the computer how much the robot is bending with every step.
The Stop Command: They added a special "pinch zone" on the robot's back. If you pinch that spot, the "Loud Ear" hears a huge signal. The robot interprets this as a stop command. If you pinch it again, it starts moving.
The Best Part: The robot has no buttons, no wires, and no external sensors. The "stop button" is just a specific spot on its own body that was printed to be extra sensitive.

Summary

The paper shows that by making the robot's skeleton out of hollow straws and carefully planning the width of those straws, you can create a robot that moves, feels, and takes commands all from a single, printed piece of material. It turns the robot's own body into its own wiring and plumbing system, eliminating the need for messy assembly.

Technical Summary: Hollow-Strut Lattice Architectures for Distributed Actuation and Sensing in Monolithic Soft Robots

Problem Statement
Monolithic soft robots fabricated via stereolithography (SLA) aim to co-locate structure, actuation, and sensing within a single printed body. However, a persistent practical bottleneck remains: establishing independent pneumatic connections to spatially distributed actuators and routing sensing channels to specific structural regions typically requires post-assembly attachment of tubing or sensing components at multiple points along the robot's surface. While architected lattice structures offer a framework for embedded routing, previous fluidic sensing approaches generally assign one channel per sensing modality, causing the number of ports and transducers to scale linearly with sensing functions. No prior lattice body has demonstrated a complete pneumatic infrastructure where the strut network itself serves as the distribution system for both actuators and multiple sensing functions through a single, geometry-programmed channel.

Methodology
The authors propose a design and fabrication strategy where the struts of an architected lattice are rendered hollow throughout, converting the scaffold itself into printed pneumatic tubing. The methodology relies on four key components:

Topology-Aware Graph Routing: A computational pipeline (implemented in Grasshopper 3D) discretizes the robot volume into a voxel grid. Passive voxels are classified as either "interface" (adjacent to actuators) or "fill." Interface voxels utilize an edge-octahedron cell topology to provide unique graph entry nodes for actuators, while fill voxels use Body-Centered Cubic (BCC) or Face-Centered Cubic (FCC) topologies based on printability (self-supporting angles) and routing capacity. A Breadth-First Search (BFS) algorithm plans two independent through-channel paths (actuation and sensing) on the lattice graph. The algorithm forbids edge reuse but allows node sharing, ensuring independent fluidic paths that terminate at a single proximal face.
Diameter-Programmed Sensitization: The local inner diameter ( $d_i$ ) of the sensing channel segments acts as a sensitivity dial. Segments printed with a 2.0 mm inner diameter (thicker walls) serve as "quiet" routing, exhibiting minimal response to deformation. Segments printed with a 2.5 mm inner diameter (thinner walls) act as sensitized zones, responding to both bending and local pinching. This allows the placement of sensing and command zones to be a design parameter rather than an assembly step.
Amplitude-Based Mode Separation: The single sensing channel is read at both ends by two pressure transducers with deliberately mismatched ranges: a high-sensitivity sensor (±100 mbar) and a wide-range sensor (±1034 mbar). Gradual structural bending produces small pressure changes detected by the high-sensitivity sensor. Sharp external pinching drives the channel pressure beyond the high-sensitivity sensor's range (saturation), generating large transients captured by the wide-range sensor. This separates deformation modes by amplitude rather than requiring separate physical channels.
Fabrication and Resin Evacuation: The entire system is printed as a single-material part using Elastic 50A resin. The "through-channel" nature of the routing allows for the flushing of uncured resin with isopropyl alcohol from one port to another, ensuring the internal channels are clear before the transducers are attached.

Key Contributions
The paper identifies four primary contributions:

Hollow Lattice Struts as Printed Tubing: A methodology that treats the lattice scaffold as the pneumatic distribution system, eliminating the need for post-assembly fluidic integration.
Diameter-Programmed Sensitization: A technique where the inner diameter of a single sensing channel defines its sensitivity profile along its length, replacing the need for dedicated per-modality routing.
Amplitude-Based Mode Separation: A dual-transducer readout strategy that distinguishes bending from tactile pinch input by exploiting transducer saturation, allowing simultaneous sensing from a single channel.
Monolithic Crawling Robot: A demonstration of a single-print robot with internal fluidic connections, capable of friction-based locomotion, bending sensing, and tactile stop commands.

Results
The architecture was verified through straight-tube characterization, passive and active lattice cubes, and a crawling robot prototype:

Sensitivity Characterization: Straight tubes with 2.5 mm inner diameters showed a sensitivity of 30 mbar/N to indentation, significantly higher than 2.0 mm (17 mbar/N) and 1.5 mm (3 mbar/N) variants. Only the 2.5 mm variant produced usable responses to bending up to 90°.
Mode Separation: In combined bending and pinching tests, the high-sensitivity sensor resolved gradual bending but saturated during pinching. The wide-range sensor captured the large transients of the pinch, confirming that bending and contact events could be separated from a single channel without additional logic.
Lattice Verification: Passive cubes demonstrated multi-channel routing capacity and independent sensing. Active cubes verified that a single sensing channel could detect bending and localized pinching simultaneously while the lattice underwent pneumatic actuation, with actuation-induced pressure changes remaining distinguishable from external tactile inputs.
Crawler Performance: The monolithic crawler (100 × 50 × 50 mm) achieved forward locomotion on a sawtooth substrate via friction-based gait, advancing at a mean velocity of 0.95 mm/s. The dorsal sensitized zone provided repeatable bending signals phase-locked to the gait cycle. A pinch on the command zone saturated the high-sensitivity sensor and triggered a transient on the wide-range sensor, successfully acting as a tactile stop and resume command.

Significance and Claims
The authors claim that this work resolves the practical limitation of requiring post-assembly fluidic integration in monolithic soft robotics. By converting the lattice scaffold into printed tubing, the external interface is reduced to a fixed set of ports at a single proximal face, regardless of the number of actuators or sensing zones. The approach extends the monolithic fabrication strategy to the fluidic connection layer itself.

The paper emphasizes that diameter-programmed sensitization changes how distributed sensing scales, as adding a sensing zone does not require additional ports or transducers. Furthermore, the dual-range readout converts the limitation of air-based sensing (lower signal amplitude and transducer saturation) into a functional mechanism for mode classification. The work demonstrates that command interfaces (such as pinch-to-stop) can be printed directly into the robot body as design outputs rather than assembled components. The authors note limitations regarding the sensitivity of air-based sensing to environmental pressure, the minimum strut wall thickness imposed by SLA resolution, and the ambiguity of single-channel localization when multiple zones are deformed simultaneously, identifying these as areas for future optimization.