Introduction
Picture a robot squeezing through a 14-millimeter gap, climbing a wall, then reshaping itself to slide under a collapsed beam. That's not science fiction — it's a 2025 lab demonstration. Meanwhile, the robots welding car bodies and tending CNC machines on factory floors today still look like they did a decade ago: rigid arms, steel joints, servo-driven precision.
That gap is the story of soft robotics. Researchers are racing to build machines from compliant materials that bend, stretch, and adapt like living tissue. Manufacturers still bet on rigid systems because cycle time, repeatability, and payload capacity decide uptime and cost per part.
This survey covers what soft robotics is, the materials and actuators behind it, how sensing and embodied AI narrow the gap with biology, and how these systems are built and deployed. It also marks where the research meets the automation manufacturers already depend on.
Key Takeaways
- Soft robots trade rigid precision for compliance, safety, and access to confined or delicate spaces
- Dielectric elastomers and Peano-HASEL actuators deliver muscle-like strain and self-healing
- Embedded sensors and onboard AI enable real-time hazard evasion and environmental tracking
- Compliant grippers and cobots already bring soft robotics into today's factories
- Industrial adoption is hybrid and component-level; fully soft robots remain lab-stage
What Is Soft Robotics? Defining a New Robotic Paradigm
Soft robotics is the design, control, and fabrication of robots built from compliant materials rather than rigid links. Instead of steel arms and motor-driven joints, these machines use elastomers, gels, and flexible composites that deform on command.
That compliance matters most in human-robot interaction. A rigid robotic arm moving at speed can cause serious injury on contact. A soft actuator, by contrast, absorbs impact rather than transmitting it. Softness here is a mechanical property, not just a design choice.
Softness isn't all-or-nothing. It shows up in two forms:
- Localized softness: soft end-effectors mounted on otherwise rigid arms, useful for handling delicate parts without redesigning the whole robot
- Full-body softness: entire robots built from compliant materials, capable of squeezing into disaster-relief rubble or navigating inside the human body
Biomimicry Drives the Design
Most soft robots borrow directly from soft-bodied organisms such as octopuses, caterpillars, and millipedes that use compliance to move efficiently through cluttered, unpredictable environments. Researchers aren't copying appearances; they're translating mechanical strategies like distributed contact points and body-wide flexibility into engineered systems.
A recent example is "FEbot," a millipede-inspired platform published in Nature Communications in 2025. It combines flexible electronic modules with arrays of directional setae (bristle-like structures) to achieve multiple locomotion modes. Those modes include vertical climbing and travel through confined passages, illustrating bio-inspired modularity in practice.
The core engineering problem: soft robots have low mechanical impedance, which makes them resistant to the rigid-body math that governs traditional robotics. Predicting how a floppy, continuously deformable structure will move under load is genuinely hard. That difficulty is pushing researchers toward automated design tools, including evolutionary algorithms, to generate and test soft robot geometries faster than manual engineering allows.
Core Materials and Actuation Methods Powering Soft Robots
Soft robots need a way to move without rigid motors, and the field has converged on a handful of actuation strategies. Each trades off voltage, force, speed, or complexity differently.
Electric-Field-Based Actuation
Dielectric elastomer actuators (DEAs) use high-voltage electric fields to squeeze and deform a soft polymer film, changing its shape. A 2024 study reported a DEA specimen reaching 253% area strain at 46 MV/m and 225 J/kg of energy density. That performance rivals biological muscle in some respects, according to research published in Nature Communications.
The catch: that kind of output typically requires field strengths above 100 MV/m, and DEAs remain vulnerable to dielectric breakdown (electrical failure under stress).
Newer designs address this directly. Peano-HASEL actuators replace solid dielectric with a liquid dielectric sealed inside flexible pouches. When voltage is applied, electrostatic force pulls the pouch closed like a zipper, displacing the liquid and producing contraction.
Early prototypes achieved 10% contraction at 50 Hz and lifted over 200 times their own weight. Because the dielectric is liquid, it re-insulates itself after minor electrical damage, giving these actuators a self-healing quality solid-film DEAs lack.
Thermal and Pressure-Based Actuation
Shape memory polymers and shape memory alloys "remember" a manufactured shape and return to it when heated. They deliver strong force relative to their mass, but they're slow: cooling, not heating, is usually the bottleneck.
Pneumatic artificial muscles work differently: a pressurized bladder inside a braided mesh expands radially and contracts axially, mimicking how biological muscle shortens. These are typically controlled with PID algorithms and require an external compressed air supply, which limits untethered use.
| Actuation Method | Strength | Key Trade-off |
|---|---|---|
| Dielectric elastomer | Very high strain, self-sensing | Needs high voltage, breakdown risk |
| Peano-HASEL | Self-healing, muscle-like contraction | Still requires kilovolt-range drive |
| Shape memory alloy/polymer | High force-to-mass ratio | Slow response, fatigue over cycles |
| Pneumatic artificial muscle | Smooth, biologically similar motion | Needs compressed air, nonlinear control |
The FEbot platform mentioned earlier combines several of these principles (super-elastic shape-memory-alloy setae paired with flexible electronics) into a hybrid, oscillation-driven system. In testing, it reached a top speed of 257 mm/s, nearly two body lengths per second, and one configuration carried a payload 10.8 times its own body mass. Those figures show soft locomotion is moving past lab demos and starting to carry measurable, useful loads.
Sensing and Embodied AI: The Next Frontier in Soft Robotics
A soft robot that can't sense its environment is just a moving blob. Soft sensors close that gap, and they fall into five main categories:
- Piezoresistive — measures deformation through resistance changes; prone to hysteresis
- Capacitive — tracks pose and contact; sensitive to nearby objects and moisture
- Optical — detects bending, force, and surface texture via light changes
- Magnetic — infers shape and position from field distortions
- Acoustic — picks up vibration and light touch; often paired with machine learning
Most of these sensors work by measuring physical deformation and inferring position or applied force from it — an indirect but effective way to give a soft body proprioception.
Electronic Skin and Multi-Modal Perception
"Electronic skin," or e-skin, takes this further by mimicking human skin's ability to register heat, pressure, and pain simultaneously. Instead of one sensor type, e-skin layers multiple modalities — pressure, temperature, humidity, even proximity — into a single flexible surface.
The FEbot research demonstrates this well. Its sensing framework pairs proprioception (tracking its own shape, posture, and curvature) with exteroception (vision, temperature, humidity, and proximity awareness), using an onboard camera, strain sensors, and thermal sensors working together.
Embodied AI Brings Decisions Onboard
Embodied AI pushes decision-making directly into the robot's hardware rather than relying on external computers. FEbot uses hyperdimensional computing (encoding sensor data as high-dimensional vectors for fast, low-power classification) to make real-time calls without a data center in the loop.
The results are tangible: the robot accelerated from 4.4 mm/s to 105 mm/s to evade an approaching hand, and it identified specific temperature thresholds (35°C and 45°C) while navigating a thermal gradient autonomously.
Results like these help explain the commercial pull. Market Research Future projects the global soft robotics market will grow from roughly $1.50 billion in 2025 to $13.97 billion by 2035, a 25% compound annual growth rate. That's a vendor forecast, not an observed result, but it signals real capital and research momentum behind these systems.
Manufacturing Techniques for Soft Robotic Systems
You can't mill or drill a squishy, continuously curved body the way you'd machine a metal bracket. Soft robots need fabrication methods built around deformable, often multi-material geometries.
Two established routes handle this today:
- Shape Deposition Manufacturing (SDM) — alternates additive and subtractive steps to build complex, multi-material structures, allowing sensors or bearings to be embedded mid-fabrication
- Smart Composite Microstructures (SCM) — laser-cuts and laminates flexible and rigid layers together, commonly used to create flexible polymer joints in hybrid soft-rigid robots
Both work, but they're slower and more manual than modern rapid-prototyping methods.
Additive manufacturing is the more scalable path. Direct ink writing extrudes shear-thinning silicone through a syringe nozzle, layer by layer, to form programmable, bioinspired actuator shapes. One documented process used a 0.41 mm nozzle at 40 psi to print pneumatic muscles and soft fingers in a single piece.
That single-piece build removes the seams that have long been a top durability failure point in soft robot fabrication. Practical gains include:
- Programmable actuator geometries without multi-step molding
- No delamination or seam failure along bonded joints
- Faster iteration than SDM or SCM for prototypes and short runs
Real-World Applications Transforming Industries
Soft robotics isn't confined to university labs anymore, though most examples remain closer to advanced prototypes than mass deployment.
Healthcare leads in maturity:
- Soft surgical manipulators like STIFF-FLOP use hydraulic actuation for minimally invasive procedures
- Textile exosuits helped stroke survivors walk faster: one study recorded a 0.14 m/s speed gain and 32 extra meters in six minutes
- MIT's magnetically steerable thread robot navigates vessel models toward clots, tested in a silicone cerebral vessel phantom
Exploration and environmental monitoring push soft materials into extremes:
- A snailfish-inspired soft robot with dielectric elastomer fins operated at 10,900 meters in the Mariana Trench
- "DraBot," a dragonfly-shaped, electronics-free silicone robot, signals water contamination with pH-reactive hydrogel and thermochromic pigment
Manufacturing is where soft robotics meets the plant floor work GLOBAL Automation Technologies supports every day:
- Soft grippers handle produce and irregular goods without bruising (strawberries, tomatoes, variable packaging)
- Food processors deploy polymer fingers where rigid grippers would crush product
- Commercial compliant grippers specify payloads up to roughly 2.2 kg for delicate or irregular parts, often without external air supply
Where Soft Robotics Meets Industrial Automation Today
Here's the honest picture: despite genuine lab breakthroughs, high-volume manufacturing still runs overwhelmingly on rigid robotic systems. Soft actuators simply can't yet match the speed, positioning accuracy, and payload capacity that a body shop or press line demands shift after shift.
That said, "softness" isn't absent from industrial floors. It's arriving in more modest, proven forms:
- Collaborative robots (cobots) that share workspace with human operators, useful for high-mix, lower-volume tending where flexibility beats raw throughput
- Compliant grippers and custom end-of-arm tooling designed to handle delicate or irregular parts without damage
GLOBAL Automation Technologies works in this space today. As a Level 5 FANUC Authorized System Integrator and the largest U.S. purchaser of FANUC robots among integrators in 2025, GLOBAL delivers proven systems for machine tending, dispensing, and painting—the work manufacturers need running reliably now, not years from now.
Those deployments include:
- Collaborative tending cells where cobots and operators share a workspace without traditional guarding
- FANUC paint robots built for hazardous spray environments
- Precision dispensing systems for seam sealing and structural adhesives
GLOBAL backs each system with turnkey integration and engineering services, plus the technical staffing to keep lines running—capabilities the company has combined since its founding in 2008.
As soft robotics and AI-assisted design tools mature toward industrial readiness, the manufacturers best positioned to adopt them will already run efficient, optimized automation. GLOBAL's simulation-driven engineering compresses robot programming timelines. Add AI-driven predictive maintenance and machine vision guidance, and that stack becomes a practical foundation for future-ready automation.
Frequently Asked Questions
What is the difference between soft robotics and rigid robotics?
Soft robots use compliant, deformable materials instead of rigid links, trading some precision and speed for adaptability and safety. Rigid robots deliver the speed and payload capacity industrial manufacturing needs.
What materials are commonly used to build soft robots?
Common materials include silicone elastomers, shape memory polymers and alloys, dielectric elastomers, and flexible electronic composites. Many designs combine several material classes into hybrid structures.
How do soft robots sense and respond to their environment?
They use flexible sensors (piezoresistive, capacitive, optical) to detect deformation, temperature, and proximity. Increasingly, embedded AI processes that data onboard to adapt behavior in real time.
Can soft robots be used in industrial manufacturing today?
Fully soft robots remain largely experimental. Soft-inspired components, like compliant grippers and cobots, are already deployed in manufacturing settings for delicate or high-mix handling tasks.
What are the biggest challenges facing soft robotics adoption?
Control complexity from low mechanical impedance and material fatigue under repeated cycling remain hard problems. Scaling lab-scale payloads to industrial speed and force requirements is still unresolved.
What is embodied AI in the context of soft robots?
Embodied AI integrates sensing, decision-making, and actuation directly into a robot's onboard hardware. This enables autonomous, real-time adaptive behavior without relying on external computing power.
