Series Elastic Actuation (SEA)

The defining feature of an SEA is a compliant element—typically a linear or torsional spring—placed in mechanical series between the motor's output (e.g., a ball screw or gearbox) and the robot's link or end-effector. This creates a low-impedance mechanical filter that:

• Absorbs impacts and shock loads, protecting gears and motors from damage.
• Stores and releases energy elastically, improving efficiency in cyclic motions like walking.
• Decouples the high-inertia rotor dynamics from the output, leading to smoother force transmission.

By measuring the deflection of the spring (via an encoder or strain gauge) and knowing its spring constant, the actuator can directly and accurately calculate the output force (τ = k * Δθ). This enables:

• Direct force control without needing a noisy joint torque sensor.
• High bandwidth force feedback for delicate manipulation and safe human-robot interaction.
• True torque control at the output, which is critical for implementing impedance and admittance control strategies that require precise regulation of the interaction dynamics.

The elastic element acts as a mechanical energy buffer. In dynamic tasks like legged locomotion, energy that would otherwise be dissipated as heat during braking can be stored as potential energy in the spring and reused for propulsion.

• Reduces peak motor torque requirements, allowing for smaller, lighter motors.
• Enables natural, spring-like gaits that mimic biological systems, lowering the Cost of Transport (CoT).
• This principle is central to models like the Spring-Loaded Inverted Pendulum (SLIP), which describes the dynamics of running and hopping.

SEAs provide inherent protection for the drivetrain. During unexpected collisions or foot strikes:

• The spring compresses, allowing the link to yield without requiring the motor to back-drive instantly.
• This limits the peak force transmitted to gears and bearings, drastically reducing wear and failure rates.
• It is a key enabling technology for dynamic locomotion over rough terrain and for robots operating in unpredictable environments where contacts are harsh and uncertain.

The compliance introduces a resonant mode (like a mass-spring system) that must be carefully managed. Key control implications include:

• Limited force bandwidth: The achievable speed of force changes is constrained by the spring's natural frequency.
• Stability concerns: High-gain position control can become unstable due to the phase lag introduced by the spring.
• Control design must explicitly model and damp this resonance, often using techniques like cascaded PID loops (inner motor position, outer spring force) or more advanced model-based controllers.

SEA is a hardware solution for achieving compliant actuation, while Impedance Control is a software control strategy. They are highly synergistic:

• An SEA provides the physical foundation and accurate force sensing needed to reliably implement impedance control laws.
• The controller can command a virtual spring-damper behavior, and the SEA's physical spring provides natural, low-impedance response at high frequencies.
• This combination is fundamental for Whole-Body Control (WBC) frameworks and safe Human-Robot Interaction (HRI).

SEA is critical for humanoid robots like Boston Dynamics' Atlas and research platforms such as MIT's Hermes and IHMC's Nadia. The series spring enables:

• Force-controlled locomotion for stable walking and running.
• Shock absorption during foot strikes and landings from jumps.
• Safe physical interaction by limiting peak contact forces during unexpected collisions.

In wearable robotics, SEA provides compliant, natural-feeling assistance. Key implementations include:

• Powered ankle-foot prostheses that store and release energy during gait, reducing metabolic cost for the user.
• Lower-limb exoskeletons for gait rehabilitation, where the elastic element allows the device to naturally yield to human movement.
• The spring acts as a built-in force sensor, enabling intuitive, impedance-based control that adapts to the user's intent.

Industrial collaborative robots from companies like Universal Robots and Franka Emika utilize SEA principles for intrinsic safety. The compliance provides:

• Inherent safety: The spring lowers effective inertia and limits peak impact forces during accidental contact with humans.
• Precise force control for delicate tasks like polishing, assembly, and electronic component insertion.
• Robustness to environmental uncertainties when interacting with unmodeled objects or misaligned parts.

SEA is a standard actuator in research quadrupeds and bipeds for investigating dynamic locomotion. Platforms like MIT Cheetah and ANYmal use it to achieve:

• Energy-efficient gaits by exploiting the spring's ability to store and return energy, similar to tendons in animals.
• High-bandwidth force control for dynamic maneuvers like galloping and traversing rough terrain.
• Direct measurement of ground reaction forces via spring deflection, which is essential for balance controllers like whole-body control.

In master-slave robotic systems, SEA enables high-fidelity force feedback. The actuator's spring:

• Provides accurate force rendering to the human operator by directly measuring interaction forces at the remote (slave) robot's end-effector.
• Improves telepresence stability by adding passive damping to the control loop.
• Allows for back-drivability, making the master device easy for the human to move when no forces are being commanded.

While powerful, SEA design involves critical compromises:

• Bandwidth Limitation: The resonant frequency of the mass-spring system limits maximum closed-loop force control bandwidth.
• Output Impedance: The spring increases the actuator's output impedance, which can be detrimental for precise position tracking tasks.
• Design Complexity: Selecting optimal spring stiffness, managing friction in the transmission, and protecting against spring failure add design overhead compared to rigid actuators.
• These trade-offs make SEA ideal for force-dominant tasks but less suitable for high-speed, stiff position control.

A control strategy that regulates the dynamic relationship between a robot's end-effector position and the contact force, making the robot behave like a programmable mass-spring-damper system. Unlike SEA, which is a hardware-based compliance, impedance control is a software algorithm that commands torques to achieve a desired dynamic behavior. It is crucial for safe human-robot interaction and handling delicate objects.

• Key Principle: Controls the ratio of force to motion (impedance).
• Contrast with SEA: SEA provides inherent, passive compliance; impedance control creates virtual compliance through active control.
• Example: A robot arm polishing a curved surface uses impedance control to maintain consistent contact force.

A force-reactive control strategy where an external force applied to the robot is measured (e.g., by a force-torque sensor) and used to generate a desired motion. It effectively controls the robot's compliance by mapping forces to velocities or positions. It is often used in conjunction with stiff, high-gear-ratio actuators.

• Key Principle: Controls the ratio of motion to force (admittance).
• Contrast with SEA: SEA measures force via spring deflection for direct force control; admittance control uses force as an input to a motion controller.
• Typical Architecture: Force Sensor → Admittance Law → Position/Velocity Controller → Stiff Actuator.
• Application: Cooperative carrying of heavy objects with a human, where the robot 'gives way' to human push/pull forces.

An advanced actuator design where the mechanical stiffness of the elastic element can be actively modulated in real-time. This allows a single actuator to switch between high-stiffness tasks (like precise positioning) and low-stiffness tasks (like shock absorption).

• Key Principle: Decouples the control of equilibrium position from mechanical stiffness.
• Relation to SEA: VSA can be seen as an evolution of SEA, adding an extra mechanism (e.g., a variable lever arm, antagonistic springs) to adjust spring constant.
• Advantage: Enables optimal impedance matching to a wider range of tasks and environments.
• Challenge: Increased mechanical complexity, weight, and control dimensionality compared to fixed-stiffness SEA.

The general capability to measure interaction forces between a robot and its environment. SEA is one specific method of achieving this. Other primary methods include:

• Base-Mounted Force-Torque Sensors: Placed between the end-effector and the final link. Provide a direct, high-bandwidth 6-axis measurement but are expensive and sensitive to extraneous loads.
• Joint Torque Sensing: Uses strain gauges on actuator output shafts or harmonic drive components. Provides direct joint-level torque feedback.
• Motor Current Estimation: Infers output torque from the motor's electrical current and a model of the gearbox. Low-cost but suffers from friction and backlash inaccuracies.

SEA's Role: Provides an intrinsic, low-cost method for direct force sensing through spring deflection, while also offering passive compliance.

A hierarchical control framework for complex robots (like humanoids) that coordinates all degrees of freedom to execute multiple tasks simultaneously—such as maintaining balance, tracking foot trajectories, and manipulating an object—while respecting physical constraints like torque limits and contact forces.

• Key Principle: Formulates control as a real-time optimization problem (often a Quadratic Program).
• Integration with SEA: The accurate, low-impedance force control provided by SEA actuators is a critical enabler for WBC. WBC algorithms compute desired joint torques or forces, which SEA can execute with high fidelity and without damaging gearboxes during unexpected contact.
• Example: A humanoid using WBC to lean and reach while keeping its Center of Pressure within the support polygon. SEA actuators allow it to exert precise forces on the environment during the reach.

Compliance (spring-like behavior) that is an inherent, unchangeable property of a robot's mechanical design. SEA is a canonical example of engineered passive compliance. This contrasts with active compliance (like Impedance Control), which is generated via software control of stiff actuators.

• Advantages of Passive Compliance:
  • Energy Storage: Springs can store and release energy, improving efficiency for cyclic tasks like running (see Spring-Loaded Inverted Pendulum model).
  • Intrinsic Safety: Absorbs impacts before the control loop can react, protecting both the robot and its environment.
  • Stability: Can improve stability in contact-rich tasks by filtering high-frequency disturbances.
• Disadvantage: The compliance is fixed by hardware and cannot be adapted online for different tasks, unlike Variable Stiffness Actuation.