Motivation: Why Soft Robots?
Conventional rigid robots excel in structured industrial environments: precise, fast, and powerful. But they struggle in three important contexts:
- Unstructured environments — cluttered domestic spaces, agricultural fields, and disaster zones require adaptability that rigid kinematics cannot efficiently provide
- Delicate object handling — grasping ripe fruit, raw meat, biological tissue, or flexible electronics requires compliant contact that rigid end-effectors damage
- Safe human proximity — a rigid robot arm in motion stores kinetic energy proportional to its mass and speed; contact with a person concentrates that energy at a small contact area
Soft robotics addresses these challenges by constructing robot bodies from continuously deformable, compliant materials — silicones (Shore A 20–60), hydrogels, textiles, and thermoplastic elastomers. Compliance is achieved mechanically rather than through active control: no torque sensor, force estimator, or impedance controller is needed. The material itself absorbs and distributes contact forces.
Soft robots draw inspiration from biological organisms that lack rigid skeletons: the octopus arm (infinite DOF, extreme dexterity), the elephant trunk (gentle grasping of diverse shapes), and the lamprey (undulatory swimming). These biological models achieve remarkable manipulation and locomotion without a single rigid joint — a capability that mechanical engineers are now beginning to replicate with engineered soft materials.
Types of Soft Actuators
The soft robotics field encompasses a diverse zoo of actuation technologies, each with distinct mechanisms:
1. Soft Pneumatic Actuators (SPA) The most mature soft actuator type. PneuNet (Pneumatic Network) actuators consist of interconnected chambers moulded into a silicone elastomer body. Inflation expands chambers asymmetrically, producing bending. Fibre-reinforced SPAs (with embedded inextensible fibres) can be programmed to bend, twist, or elongate depending on fibre angle.
2. Tendon-driven Soft Manipulators Inextensible cables (tendons) routed through flexible tubes or along flexible backbones. Pulling a tendon shortens the corresponding path, bending the backbone. Continuum sections can be stacked to produce hyper-redundant manipulators. Examples: Festo Bionic Cobot, endoscopic robots.
3. Dielectric Elastomer Actuators (DEA) An elastomer film (e.g., VHB acrylate or silicone) sandwiched between compliant electrodes. Applying high voltage (1–5 kV) generates Maxwell stress that compresses the film in thickness and expands it in area — producing strains of 30–300%. No pneumatic supply required; electrically driven; fast (> 100 Hz); but high voltage presents safety concerns and film pre-stretch is required.
4. Hydrogel Actuators Crosslinked polymer networks that absorb or expel water in response to stimuli: pH, temperature, moisture, or light. Differential swelling between bilayer structures generates bending. Bioinspired by pine-cone opening; used in microfluidics and untethered micro-robots. Very slow (seconds to minutes); biocompatible.
5. Fibre-Reinforced / McKibben PAMs See Chapter 17. Embedded in soft bodies these form the basis of most pneumatic exosuits and biomimetic legs.
Soft Actuator Technology Comparison
| Type | Actuation Principle | Strain Range | Blocking Force | Maturity |
|---|---|---|---|---|
| PneuNet SPA | Pneumatic chamber inflation | 50–200% bending arc | Low–medium (< 10 N at tip) | High — commercial grippers available |
| McKibben PAM | Braided mesh + bladder inflation | 20–30% contraction | High (100–1000 N) | High — commercial products (Festo Fluidic Muscle) |
| Dielectric Elastomer (DEA) | Electrostatic Maxwell stress | 30–300% area expansion | Low–medium | Medium — lab prototypes, few products |
| SMA wire | Thermally driven phase transformation | 4–8% linear contraction | High (200 MPa stress) | Medium — used in commercial grippers |
| Hydrogel | Osmotic swelling/pH response | 100–1000% volumetric | Very low | Low — research stage |
| Twisted coil (nylon) | Thermally driven coil shortening | 10–50% contraction | Low–medium | Low–medium — research demonstrations |
Soft Actuators vs Rigid Actuators
Soft Actuators
- Inherent compliance — no active impedance control needed for safe contact
- Continuous deformation enables shape adaptation to irregular objects
- Low mass and simple fabrication via moulding or 3D printing
- Force output typically lower (< 100 N for pneumatic soft actuators)
- Position accuracy limited — no rigid kinematic chain; modelling is complex (hyperelastic FEM)
- Durability limited by elastomer fatigue, especially at high cycle counts
- Control is challenging: non-linear behaviour, large deformations, model uncertainty
Rigid Actuators
- High stiffness and precision — position repeatability to < 10 µm in industrial systems
- High force and torque output — kilonewtons for hydraulic, hundreds of N·m for electric
- Well-established modelling (Denavit-Hartenberg, URDF), simulation, and control tools
- Impact hazard — kinetic energy concentrated in rigid links; requires active collision detection
- Limited environmental adaptability — fixed kinematic structure cannot conform to objects
- High complexity and cost for high-DOF systems (many motors, sensors, drivers)
- Heavy — motors, gearboxes, and structural members add mass at each joint
Key Design Challenges for Soft Actuators
- 01
Fabrication — soft lithography, moulding, and multi-material 3D printing enable complex geometries but require careful material selection (Shore hardness, tear strength, gas permeability) and quality control for embedded channel integrity
- 02
Modelling and simulation — hyperelastic material models (Neo-Hookean, Mooney-Rivlin, Yeoh) and finite element analysis are needed for design; real-time control models must be simplified (piecewise-constant curvature, neural network surrogates)
- 03
Sensing integration — embedding strain sensors, contact sensors, or proprioceptive elements without compromising compliance or fabrication simplicity; soft strain gauges (liquid metal channels, carbon black composites) are promising
- 04
Durability and fatigue — elastomers under cyclic pneumatic or mechanical loading degrade through crack propagation at stress concentrations; design of radii, wall thickness, and surface finish critically affects actuator lifetime
- 05
Control — highly non-linear input-output relationships, model uncertainty, and large deformations make closed-loop position control difficult; data-driven approaches (reinforcement learning, Gaussian processes) are active research areas