Biomimetic Actuators, Artificial Muscle Technology and Wearable Actuators — Robotics Tutorials

Biological Muscle as the Performance Benchmark

Before designing artificial muscles, engineers must understand the extraordinary performance of the benchmark: human skeletal muscle.

Skeletal muscle is a hierarchical, fibre-based actuator. At the nanoscale, actin and myosin protein filaments slide past each other in a ratchet-like mechanism powered by ATP hydrolysis. These sarcomeres (~2 µm long) are arranged in series and parallel to form myofibrils, then muscle fibres, then fascicles, and finally a complete muscle belly enclosed in connective tissue.

Key performance metrics of skeletal muscle:

Isometric stress: ~0.1–0.3 MPa (force per unit cross-section)
Strain: ~20–40% contraction from rest length
Specific power: ~50–200 W/kg (muscle mass) during fast contractions
Efficiency: 20–35% (thermodynamic; much of the rest is heat)
Bandwidth: up to ~100 Hz for fast-twitch fibres during brief tetanus
Self-healing: minor damage repaired by satellite cells within days to weeks
Proprioception: muscle spindles and Golgi tendon organs provide intrinsic length and force sensing
Scalability: same basic mechanism operates in organisms from insects to blue whales

No single artificial actuator matches all these properties simultaneously. Most exceed biological muscle in specific stress (SMA: 200 MPa vs 0.3 MPa) but fall short in strain, efficiency, self-healing, and integrated sensing. The artificial muscle challenge is fundamentally one of material integration and multi-functional design.

Artificial Muscle Technologies vs Biological Muscle

Technology	Stress (MPa)	Strain (%)	Efficiency (%)	Speed	Maturity
Biological muscle	0.1–0.3	20–40	20–35	High (to 100 Hz)	N/A — the benchmark
Hydraulic actuator	~10–35	> 50 (linear)	70–90	Very high (> 100 Hz)	Very high
Pneumatic PAM (McKibben)	~0.1–0.7	20–30	40–60	Medium (1–10 Hz)	High
SMA (NiTinol)	~100–200	4–8	< 5	Low (< 2 Hz)	Medium
Dielectric Elastomer	~0.1–1	30–300 (area)	60–80	High (> 100 Hz)	Medium
IPMC	~0.1–0.5	> 90° bend	< 5	Low (< 10 Hz)	Low–medium
Twisted coil (TCP)	~10–60	10–50	25–35	Low–medium	Low–medium
Hydrogel	~0.001–0.1	100–1000 (volumetric)	< 5	Very slow	Low

Biomimetic Actuation Strategies

Biological organisms achieve superior performance not only through better actuator materials but through architectural strategies that engineers are now applying to robotic systems:

1. Agonist-Antagonist Pairs Biological joints are driven by opposing muscle groups: the bicep flexes the elbow, the tricep extends it. No single muscle can push; it can only pull. Robotic joints implementing this architecture use two actuators (motors, PAMs, or cables) in opposition. The key advantage is variable stiffness: by co-contracting both actuators, the joint becomes stiffer; by modulating the balance, position changes. This is the principle behind the Variable Stiffness Actuator (VSA), used in the DLR Hand Arm System and Pisa/IIT SoftHand.

2. Distributed Actuation Biological muscle fibres are recruited in groups (motor units) in response to load, enabling smooth force gradation and local fatigue management. Distributed cable-driven or PAM-driven robots mimic this by actuating a continuum body with many inputs, enabling rich shape spaces from redundant actuation.

3. Tendon Routing and Cable-Driven Hands The human hand has 27 bones but only 6 intrinsic muscles in the palm and 18 forearm muscles connected by long tendons. Cable-driven robotic hands (Shadow Dexterous Hand, Allegro Hand) replicate this architecture: motors located in a remote forearm unit route Bowden cables to finger joints. Benefits include remote mass distribution, compliance via cable elasticity, and dexterity from underactuation.

4. Variable Stiffness Actuators (VSA) VSAs modulate mechanical impedance during task execution — stiff during precision positioning, compliant during impact or human contact. Designs include: dual-motor with spring coupling (qbMove), cam-based spring preload (VS-Joint), and non-linear spring mechanisms (AMASC).

Wearable Actuators: Exoskeletons and Exosuits

Wearable robotic actuators form the foundation of exoskeletons and exosuits — devices worn on the human body to augment, assist, or rehabilitate movement. Two broad paradigms have emerged:

Rigid Exoskeletons Rigid exoskeletons (Ekso Bionics EksoGT, ReWalk Robotics P6, Cyberdyne HAL) use rigid aluminium or carbon-fibre frames aligned to the user's skeletal segments, with motorised joints (DC servo or hydraulic) at hip, knee, and ankle. They can deliver large joint torques (> 100 N·m) and are used for spinal cord injury rehabilitation, industrial worker assistance, and military load augmentation.

Challenges include: precise kinematic alignment (human joint axes shift during movement, unlike fixed robot joints), donning/doffing complexity, mass (8–25 kg), and high cost (USD 70 000–150 000).

Soft Exosuits The Harvard Biodesign Lab and collaborators pioneered soft exosuits: lightweight (~1 kg) garments using Bowden cable transmissions driven by actuators mounted at the waist or back. Cables route along the outer surface of limbs and attach at anchor points on the foot, shank, and thigh. Biomechanically, the suit augments ankle push-off and hip flexion at the appropriate gait phases, reducing metabolic cost of walking by 7–23% in healthy subjects and 15–30% in post-stroke patients.

Assistive hand orthoses use pneumatic soft actuators (glove-mounted PneuNet or cable-driven fingers) to assist grasping in spinal cord injury and stroke patients, providing controlled finger extension and flexion during rehabilitation.

Rigid Exoskeleton vs Soft Exosuit

Rigid Exoskeleton

Mass: 8–25 kg — significant metabolic penalty to carry the device itself
Force/torque: very high (> 100 N·m at joints) — enables full body weight support
Compliance: low — rigid structure can resist natural joint motion if misaligned
Kinematic alignment: critical — misalignment causes painful hyperstaticity and skin pressure
Donning time: 5–20 minutes with carer assistance
Cost: USD 70 000–150 000 for medical-grade devices
Use case: complete lower-limb paralysis, heavy industrial load carrying

Soft Exosuit

Mass: 0.5–3 kg — low metabolic overhead; garment-like form factor
Force/torque: moderate (10–50 N·m) — assists gait rather than replacing muscle
Compliance: high — cable transmissions and fabric adapt to natural limb motion
Kinematic alignment: not required — cables transmit force along limb surface
Donning time: 1–5 minutes independently for trained users
Cost: USD 5 000–50 000 — lower fabrication complexity
Use case: gait assistance in stroke, CP, aging; metabolic augmentation in healthy users

Tendon-Driven Underactuated Robot Hands

The most dexterous robotic hands in existence use a tendon-driven underactuated architecture inspired directly by human hand anatomy.

In an underactuated hand, the number of actuators N is less than the number of joints M it drives (N < M). Tendon routing through pulleys, sheaths, and differential mechanisms distributes a single motor's motion across multiple finger joints. When the hand contacts an object, the contact forces constrain some joints while others continue to close — the finger conforms to the object shape automatically, without requiring complex contact sensing or planning.

Examples:

Shadow Dexterous Hand — 20 DOF, 20 tendons (near-isomorphic with human hand), driven by 40 pneumatic muscles; the most anatomically faithful robot hand
Allegro Hand — 16 DOF, 16 independent torque-controlled joints (NOT underactuated), fully actuated with brushless motors
Pisa/IIT SoftHand — 19 DOF driven by a single motor through an adaptive synergy; can grasp objects of widely varying shape through passive mechanical compliance
RBO Hand 3 — soft pneumatic fingers mounted on a 3D-printed palm; each finger actuated independently with low-pressure air

The insight from human neuroscience is that hand movements lie predominantly in a low-dimensional synergy space: the first two or three principal components of hand configuration capture > 80% of natural grasping variance. Underactuated hands exploit this by encoding common synergies mechanically, requiring fewer actuators and simpler control strategies.

Key terms

Artificial muscle: An actuator technology designed to replicate key properties of biological skeletal muscle: large strain, moderate stress, compliance, and distributed force generation.
Variable stiffness actuator (VSA): An actuator that can actively modulate its mechanical impedance, enabling both stiff precision positioning and compliant safe interaction within the same joint.
Agonist-antagonist: A paired actuation architecture in which two opposing actuators drive a joint in opposite directions, enabling variable stiffness through co-contraction.
Exoskeleton: A wearable robotic device with rigid structural frame and motorised joints that augments or replaces the wearer's muscle function.
Exosuit: A lightweight, compliant wearable robot using cable transmissions routed along a garment to assist human movement with minimal mass penalty.
Tendon-driven: An actuation architecture that uses inextensible cables (tendons) routed through sheaths to transmit force from remotely located motors to joints or end-effectors.
Underactuated: A robotic system where the number of independent actuators is fewer than the number of degrees of freedom; exploits mechanical constraints for passive adaptation.
Power density: Mechanical power output per unit mass (W/kg); biological muscle achieves ~50–200 W/kg, a target for artificial muscle development.
Biomimetic: Inspired by or modelled on biological structures, mechanisms, or processes; biomimetic actuators replicate architectural strategies found in living organisms.
Series elastic actuator (SEA): An actuator with a compliant spring element in series between the gearbox and the load, providing accurate torque sensing and passive impact absorption.

← Actuators & Motion