How soft robots are learning to feel their way through the real world

Robots are often imagined as rigid metal machines, all gears and bolts. Yet a rapidly growing field called soft robotics is reimagining them as flexible, squishy and almost creature like. These new designs promise safer machines that can better handle delicate tasks and adapt to messy, unpredictable environments.

From surgical tools that gently grip tissue to flexible grippers that can sort fragile fruit, soft robots are starting to leave the lab. Scientists are now racing to give them something more: a sense of touch and awareness of their own shape.

What makes a robot “soft”

Soft robots are built from materials such as silicone, rubber, textiles and gels instead of rigid metals. Many are powered by air or fluid pumped into internal chambers, which bend and stretch like muscles when inflated. Others use stretchable electrical components or shape changing polymers.

This softness has a clear advantage. If a rigid robot arm bumps into a person, the impact can be dangerous. A soft arm can absorb some of the force and deform around obstacles. That makes it attractive for work alongside people in warehouses, homes and hospitals.

Why sensing is so hard for flexible machines

Traditional industrial robots rely on precise joints, motors and encoders that measure angles. Software can easily calculate the position of each rigid link, so the machine always “knows” where its parts are in space. For a soft robot that can bend and twist anywhere along its body, this approach breaks down.

To move reliably, a soft robot needs to sense its shape in real time, much like an octopus arm or a human finger. It also needs to feel contact forces so that it can gently grasp objects without crushing them. Achieving this requires new kinds of sensors that stretch, compress and flex without breaking.

New materials that feel pressure and stretch

Researchers are developing flexible sensors that can be embedded directly into soft structures. One common design uses thin channels filled with liquid metal or conductive fluid. When the robot bends, the channel geometry changes and so does its electrical resistance, which can be measured and translated into shape information.

Another approach relies on soft capacitive sensors, where two flexible conductive layers act like plates of a capacitor separated by an elastic material. When pressed or stretched, the distance between the plates changes, altering the electrical signal. These thin, skin like layers can cover a robot arm, giving it a distributed sense of touch similar to human skin.

Teaching soft robots to interpret their own bodies

Sensing is only half of the challenge. The robot must also interpret the flood of data that comes from dozens or even hundreds of embedded sensors. Machine learning is becoming an important tool for this task, especially techniques that map raw signals to approximate shapes and contact patterns.

In some experiments, scientists train models by bending a soft robot through many shapes while recording sensor outputs and precise motion capture data. Once trained, the robot can estimate its own posture from sensor readings alone, without external cameras. This kind of “proprioception” lets it move with more confidence in cluttered spaces or underwater environments.

Everyday applications: from fruit picking to delicate surgery

As sensing improves, soft robotic grippers are finding practical uses in agriculture and logistics. Flexible fingers can conform around irregularly shaped fruit, pastries or containers while built in pressure sensors ensure that the grip stays firm yet gentle. This reduces bruising and damage, a major concern when handling food or fragile goods.

In medicine, soft robotic tools with tactile feedback could improve minimally invasive procedures. A surgeon controlling a flexible instrument might feel subtle differences between healthy and diseased tissue through haptic interfaces. Better sensing could help avoid accidental damage and enable more precise interventions deep inside the body.

Challenges on the way to robust soft machines

Despite rapid progress, soft robotics faces practical hurdles. Soft materials can tear, fatigue and degrade more quickly than rigid metals, especially when stretched repeatedly. Sensors must survive these strains while maintaining accuracy over time, which is difficult for delicate conductive paths and gels.

Power and control are also open problems. Many prototypes rely on bulky external pumps or wires that limit mobility. Researchers are experimenting with compact pumps, embedded batteries and even chemical actuators that release energy within the material itself. Integrating sensing, power and control into a single, robust body remains a central engineering challenge.

Why soft robots matter for the future of technology

Soft robots sit at the crossroads of materials science, biology, physics and control engineering. Their development is not just about building novel gadgets, it is about rethinking how machines interact with the real world and with people. By blending flexibility, sensing and intelligence, they offer a path toward robots that are less like rigid tools and more like adaptive partners.

Many of the underlying technologies, such as stretchable electronics and tactile sensors, can also feed into other products from wearable health monitors to smarter prosthetic limbs. As these components become more reliable and affordable, the line between “robotic” and everyday devices may start to blur in surprising ways.