Soft robotics is a branch of robotics that builds machines from highly deformable materials, such as elastomers, gels, and flexible polymers, rather than the rigid metals and hard plastics that make up conventional robots. These materials are chosen specifically because they can bend, stretch, and compress in ways that resemble biological tissue and organs.1Soft Robotics. Soft Robotics: A Perspective—Current Trends and Prospects for the Future The result is a class of machines that can squeeze through tight gaps, conform to irregular shapes, and interact safely with fragile objects or human bodies. How these machines actually generate movement, sense their surroundings, and get told what to do involves a set of engineering strategies quite different from anything in traditional robotics.
What Makes a Robot “Soft”
A conventional industrial robot arm is essentially a series of stiff links connected by joints, like a skeleton. It moves precisely, but it is also unyielding. If it bumps into a person or a delicate object, something breaks. A soft robot, by contrast, is built from materials that deform easily under stress and spring back when the stress is removed. Silicone rubber (particularly the family of materials called polydimethylsiloxane, or PDMS) is the workhorse material of the field. Other common choices include hydrogels, polyurethane foams, liquid-metal-embedded elastomers, and various polymer composites.2Materials Today: Proceedings. A review on soft materials utilized for the manufacturing of soft robots
What all these materials share is low stiffness. Where steel has a stiffness measured in the hundreds of gigapascals, the elastomers used in soft robots sit in the kilopascal-to-low-megapascal range, roughly matching the mechanical properties of skin, muscle, and cartilage. That match is not a coincidence. Much of the appeal of soft robotics lies in building machines that can touch, grip, and move alongside living things without causing harm.
Fabrication has shifted dramatically in recent years. Early soft robots were made with conventional silicone molding, but three-dimensional printing now allows researchers to build complex internal channels and layered structures that would be impossible to mold. A technique called direct ink writing extrudes silicone or hydrogel inks through a nozzle to build up structures layer by layer, and it has become a new standard for producing the tiny internal channels that many soft robots rely on for movement.3MRS Communications. Direct ink writing of silicone elastomers to fabricate microfluidic devices and soft robots Hydrogel-based robots can also be 3D printed using biocompatible alginate as a thickening agent, opening doors to robots meant to work inside the body.4PubMed. Direct-Ink-Write 3D Printing of Hydrogels into Biomimetic Soft Robots
How Soft Robots Move
Making a squishy object do useful work is the central engineering challenge of the field. Conventional motors and gears assume rigid components. Soft robots need actuation strategies that work with deformable bodies, and several approaches have emerged.
Pneumatic Networks
The most widely used method pumps air into tiny channels molded or printed inside the robot’s elastomeric body. These channels are arranged in patterns called pneumatic networks, or “pneu-nets.” When air pressure increases, the channels inflate, and because the walls are designed with varying thickness or stiffness, the inflation produces controlled bending, twisting, or extending movements.5Advanced Functional Materials. Pneumatic Networks for Soft Robotics that Actuate Rapidly The appeal is that you get sophisticated motion from very simple controls: pump air in, the robot curls; release it, and the robot straightens.6PubMed. Analytical Modeling and Design of Generalized Pneu-Net Soft Actuators with Three-Dimensional Deformations Most soft grippers you see in demonstrations use some version of this concept. The downside is that you need an air supply, which is bulky. Tethered pneumatic lines work fine in a lab or a factory, but they limit portability.
Electrically Driven Materials
A second major approach uses materials that change shape when an electric field is applied. Dielectric elastomers are thin rubber films sandwiched between flexible electrodes. Apply voltage and the film compresses in thickness and expands in area, producing motion. These actuators can deform substantially, respond quickly, and pack a lot of energy into a small volume.7Advanced Intelligent Systems. Review of Dielectric Elastomer Actuators and Their Applications in Soft Robots A broader family of materials called electroactive polymers includes ionic variants that move when ions migrate through them, offering lightweight and energy-efficient actuation that researchers compare to artificial muscles.8PubMed Central. Research Progress in Electroactive Polymers for Soft Robotics and Artificial Muscle Applications Electric actuation avoids the bulky compressors needed for pneumatics, but the voltages involved can be high, and durability remains a concern.
Shape Memory Materials and Thermal Actuation
Some soft robots use materials that “remember” a pre-programmed shape and return to it when heated. Shape memory alloys (thin metal wires that contract when heated above a threshold temperature) and shape memory polymers (plastics that soften and reform when warmed) can both serve as internal muscles. A recent design combines both in a single structure: alloy springs generate the movement, while the polymer core locks the shape in place once the robot reaches its target configuration, and also controls how stiff the robot is at any given moment by regulating temperature.9Smart Materials and Structures. Programmable structure with shape memory materials for soft robotics This kind of adjustable stiffness is valuable because it lets a robot be compliant when it needs to squeeze through a gap and rigid when it needs to push against something.
An emerging twist on thermal actuation uses 4D printing, where a 3D-printed object is designed to change shape over time in response to a stimulus. Magnetically responsive shape memory polymers, for instance, can be printed into structures that reshape themselves when exposed to an external magnetic field and heat.10PubMed Central. 4D Printing of Magnetically Responsive Shape Memory Polymers: Toward Sustainable Solutions in Soft Robotics, Wearables, and Biomedical Devices This opens the door to robots that can be manufactured flat and then “activated” into their functional shape on demand.
Biohybrid Approaches
At the more experimental end of the spectrum, some researchers are building robots that incorporate living cells. Biohybrid soft robots combine synthetic scaffolds with living muscle tissue, creating machines that use biological contraction as their power source. These devices, still at laboratory scale, include tiny swimmers, crawlers, and walkers whose movement comes from cultured muscle cells responding to electrical or chemical stimulation.11PubMed. Biohybrid soft robots with self-stimulating skeletons Living tissue offers responsiveness and adaptability that synthetic materials cannot yet match, though keeping biological components alive and functional outside of laboratory conditions remains a serious hurdle.
How Soft Robots Sense Their World
A robot that can bend and stretch is not much use if it cannot tell where its own body is or what it is touching. Sensing in soft robotics is a distinctly tricky problem because the sensors themselves need to deform along with the body. A rigid strain gauge glued to silicone would crack or pop off the first time the robot flexed.
Flexible and stretchable pressure sensors have advanced rapidly in response to this need. These devices use mechanisms like changes in electrical resistance or capacitance when pressure is applied, and they can monitor forces ranging from the gentle pulse of a heartbeat to the firm grip of a robotic hand.12PubMed Central. Flexible and Stretchable Pressure Sensors: From Basic Principles to State-of-the-Art Applications One persistent challenge has been that stretching a pressure sensor often scrambles its pressure readings, since both stretching and pressing deform the material. A recent design solved this by using an ionic capacitive mechanism combined with a hierarchical microstructure, achieving pressure sensitivity that stayed stable even when the sensor was stretched to half again its original length. That sensor could detect pressures as light as 0.2 pascals, which is roughly the force of a breath on your skin.13PubMed Central. A stretchable and strain-unperturbed pressure sensor for motion interference-free tactile monitoring on skins
Beyond touch, soft robots also need proprioception, the ability to know where their own body parts are in space. Traditional robots solve this trivially with encoders at each joint, but a soft robot has no joints. One recent approach embeds strain sensors along the length of a slender soft robot and uses a modeling framework to reconstruct the robot’s full 3D shape from those strain readings, achieving high accuracy even in complex bent configurations.14PubMed. Advancing Soft Robot Proprioception Through 6D Strain Sensors Embedding Getting proprioception right is essential for closed-loop control, where the robot adjusts its behavior based on real-time feedback rather than just executing pre-programmed moves.
Why Controlling a Soft Robot Is Hard
If you push on a rigid robot’s joint, you know exactly what will happen because the link between input and output is stiff and predictable. Push on a soft robot and the whole body deforms in complicated, nonlinear ways. The same input pressure might produce different movements depending on whether the robot is resting on a table, dangling in the air, or pressing against an obstacle. This makes traditional control engineering, which relies on precise mathematical models of how a system responds, extremely difficult to apply.
Researchers are increasingly turning to machine learning to bridge the gap. Deep reinforcement learning, for example, lets a simulated soft robot try thousands of movement strategies, learn which ones work, and then transfer that knowledge to a physical robot. One team used this approach to train a soft robotic arm to perform dynamic pushing tasks, building a physics model of the robot’s flexible body in simulation and then refining the controller with real-world feedback.15Advanced Intelligent Systems. Pushing with Soft Robotic Arms via Deep Reinforcement Learning
A related concept gaining traction is “embodied intelligence,” the idea that the soft body itself does part of the computation. When an octopus arm wraps around an irregular object, the arm’s compliance automatically distributes grip force without the brain having to calculate anything. Soft robots can exploit the same principle: a well-designed gripper conforms to a tomato’s shape passively, without needing a camera or a pressure-control algorithm to map the surface.16PubMed Central. Exploring Embodied Intelligence in Soft Robotics: A Review The physical form and material properties of the robot handle part of what would traditionally be a software problem. The field is still working out how to combine this “body intelligence” with learned controllers to get the best of both.
Nature as the Design Guide
Soft robotics borrows heavily from biology, and not just in the abstract sense that “nature inspires us.” Specific animals and plants serve as detailed blueprints. Octopuses, with their boneless arms, are the iconic example, but crawling caterpillars, slithering snakes, and climbing geckos have all shaped specific designs.17PubMed Central. Caterpillar-inspired soft crawling robot with distributed programmable thermal actuation A caterpillar-inspired crawler, for instance, uses patterned silver nanowire heaters embedded in a liquid crystal elastomer to mimic the wave-like contractions a real caterpillar uses to inch forward, achieving multiple crawling modes with a single material system.
Plants offer their own inspiration. Tendrils that coil around supports perform two distinct motions, bending and helical twisting, triggered by environmental cues like sunlight or humidity. Researchers have replicated both motions in a dual-layer liquid crystal actuator that bends under ultraviolet light and twists under near-infrared light, switching between the two behaviors simply by changing the wavelength of illumination.18PubMed Central. A plant tendril mimic soft actuator with phototunable bending and chiral twisting motion modes This is a useful reminder that “soft robotics” is not just about squishy grippers. It also encompasses light-driven, shape-shifting structures that respond to their environment without any electronic control at all.
Where Soft Robots Are Already Working
The field has moved well past proof-of-concept demonstrations, though many applications are still in early deployment. Three areas stand out for their practical momentum.
Medicine
Healthcare is where soft robotics may have the most transformative impact. Applications span from surgical tools that can navigate the twisting corridors of the gastrointestinal tract, to soft implants, to wearable exosuits that assist people recovering from strokes or spinal injuries.19PubMed Central. Pioneering healthcare with soft robotic devices: A review One striking example is a soft robotic sleeve designed to wrap around a failing heart. The sleeve compresses and twists in sync with the heart’s natural rhythm, acting as a ventricular assist device. Unlike traditional mechanical heart pumps, which contact the blood and carry clotting risks, this sleeve works from the outside.20PubMed. Soft robotic sleeve supports heart function The concept of an implant that matches the softness of the organ it supports is a fundamentally different design philosophy from anything in conventional biomedical engineering.
Agriculture and Food Handling
Picking a ripe strawberry without crushing it is easy for a human hand but nearly impossible for a rigid robotic gripper. Soft grippers have emerged as a promising solution for handling produce and other fragile items in agriculture.21Advanced Intelligent Systems. Intelligent Soft Robotic Grippers for Agricultural and Food Product Handling: A Brief Review with a Focus on Design and Control These grippers conform to irregular shapes naturally, distributing contact pressure so that soft fruit is not bruised. The same principle applies to automated harvesting, where the gripper must deal with unpredictable object shapes, sizes, and positions on the plant.22PubMed Central. Soft Grippers for Automatic Crop Harvesting: A Review This is the embodied intelligence idea in action: the gripper does not need to identify the exact shape of every tomato because its compliant fingers adapt on contact.
Search and Rescue
After earthquakes, building collapses, or mine accidents, finding survivors in tight rubble requires something that can navigate spaces too confined and unpredictable for a rigid wheeled or tracked robot. Soft robots can squeeze through narrow gaps, change shape to navigate around obstacles, and even crawl across mixed terrain. Bio-inspired locomotion strategies like crawling, rolling, and shape morphing give these robots capabilities that rigid platforms simply cannot replicate.23PubMed Central. An Amphibious Fully-Soft Centimeter-Scale Miniature Crawling Robot Powered by Electrohydraulic Fluid Kinetic Energy Some prototypes are fully soft and centimeter-scale, able to operate in both wet and dry environments, which is valuable in flooded disaster zones.
Safety Around People
One of the strongest practical arguments for soft robotics is safety during physical contact with people. Industrial robots have historically needed cages or exclusion zones because a collision with a rigid arm moving at speed can cause serious injury. Collaborative robots (cobots) have started sharing workspace with people, but they must adhere to strict safety standards that limit their speed, force, and power. The international technical specification ISO/TS 15066 defines allowable contact forces and pressures for different body regions during human-robot interaction.24Results in Engineering. Physical human-robot interaction: A critical review of safety constraints
Soft robots offer a fundamentally different approach to this problem. Instead of limiting the robot’s speed and force through software and sensors, the robot’s own body absorbs and distributes impact energy. A collision with a silicone arm at moderate speed feels more like bumping into a pool noodle than getting hit by a metal bar. This inherent compliance does not eliminate the need for safety engineering, but it lowers the baseline risk enough to open up applications where rigid cobots would be impractical or where the safety restrictions would make them too slow to be useful. Wearable rehabilitation devices, assistive exosuits, and prosthetics all benefit directly from this property.
Sustainability and End-of-Life Concerns
As soft robotics scales up, the environmental footprint of its materials is drawing attention. Silicone elastomers are durable, but they are not readily biodegradable, and most soft robots today end up as waste at the end of their useful life. Some researchers are tackling this head-on. One team developed electrically conductive coatings based on a biodegradable vitrimer binder combined with graphene and carbon nanofibers. These coatings can be sprayed onto soft robot bodies to serve as flexible wiring or sensor layers, and they are designed to be self-healing, recyclable, and eventually biodegradable.25Advanced Sustainable Systems. Self‐Healing, Recyclable, Biodegradable, Electrically Conductive Vitrimer Coating for Soft Robotics Self-healing is a particularly interesting property: if a sensor crack or a wiring break can repair itself, the robot lasts longer before it becomes waste in the first place.
The broader push in the field is toward materials that are recyclable or bio-derived without sacrificing performance. Hydrogel-based robots, for example, are largely water and can dissolve harmlessly at end of life, but they are mechanically weak and dry out in open air. Finding the right balance between performance, durability, and environmental responsibility is an active area of materials research.
The Power Problem
For all their promise, soft robots face a persistent limitation that rarely makes it into popular accounts of the field: they struggle to produce high output power. Conventional rigid robots are strong because stiff structures transmit force efficiently. Soft materials absorb and dissipate energy rather than transmitting it, and their actuation mechanisms tend to be slower than electric motors. Energy storage is limited, response times can lag, and a significant fraction of input energy is lost as heat or internal friction.26Fundamental Research. Actuation strategies and applications for high-output power soft robots
This means soft robots currently excel at tasks requiring gentle manipulation, conformability, and safe interaction, but they are poor candidates for jobs requiring heavy lifting or fast, forceful motion. A soft gripper can pick a strawberry beautifully, but it cannot move a cinder block. Addressing the power gap will likely involve hybrid designs that combine soft exteriors with stiffer internal structures that can be engaged when more force is needed, as well as new actuation materials that can convert electrical or chemical energy into mechanical work more efficiently. For now, though, the power ceiling is probably the single biggest factor shaping where soft robots can and cannot be deployed.