Series Elastic Actuator: Design and Control Guide

Introduction

A series elastic actuator places a deliberately compliant element between a motor transmission and its load, turning spring deflection into a practical measure of output force. This guide explains its mechanical layout, governing equations, control strategy, and value in robots that must move dynamically while interacting safely with people or uncertain environments.

How a Series Elastic Actuator Works

A conventional rigid actuator connects its motor and gearbox as stiffly as possible to the joint. In a series elastic actuator, a calibrated linear or torsional spring sits in the load path, while encoders measure motor position and joint position. Their relative displacement reveals how far the spring has deformed.

This compliance performs two jobs. It absorbs impact energy before a shock reaches fragile gears, and it acts as a mechanical force sensor. The controller can therefore regulate interaction torque without relying only on motor current, which is distorted by gearbox friction, backlash, and changing efficiency.

The trade-off is reduced mechanical bandwidth: a soft spring improves sensitivity and impact tolerance but allows more joint deflection. A stiff spring gives quicker position response but produces smaller sensor signals and less shock isolation. Selecting spring stiffness is consequently a system-level decision involving load, speed, sensing resolution, and control frequency.

Series Elastic Actuator Torque and Spring Stiffness

For a rotary compliant actuator operating in its linear range, transmitted torque is τ = kθ, where τ is torque in N·m, k is torsional spring stiffness in N·m/rad, and θ is relative angular deflection in radians. For a linear design, the equivalent relation is F = kx. These equations make force sensing direct, provided the spring is calibrated and hysteresis remains small.

Consider an exoskeleton knee requiring 30 N·m of assistance. If its torsional spring stiffness is 600 N·m/rad, the required deflection is θ = τ/k = 30/600 = 0.05 rad, or about 2.86°. An encoder must resolve this deflection accurately across the complete load range.

Torque control normally uses the measured spring torque as feedback. The controller compares desired torque τd with measured torque τm, forms error e = τd − τm, and commands motor current or velocity through a PI or PID loop. Engineers must also check the spring’s maximum stress, fatigue life, resonant frequency, and stored energy U = 0.5kθ².

Applications in Humanoid Robot Actuators and Exoskeletons

Humanoid robot actuators benefit from compliance during walking, landing, grasping, and accidental collisions. At an ankle or knee, the spring can temporarily store energy and smooth peak loads; at an arm, torque control lets the robot press, polish, or handle an object without imposing rigid motion. These properties support physical AI systems whose software must act through real mechanisms.

Powered prostheses and exoskeleton robotics use similar principles because contact forces directly affect the wearer. Series elasticity can make assistance feel less abrupt while providing measurable joint torque for gait control. Other applications include rehabilitation robots, quadrupeds, teleoperation devices, and force-controlled assembly.

Compliance does not automatically make a machine safe. Designers still need mechanical stops, torque limits, emergency circuits, suitable guarding, and risk assessment. The spring improves intrinsic behavior, but the complete robot determines safety.

Common Series Elastic Actuator Mistakes and Exam Tips

A common design error is choosing spring stiffness from peak torque alone. Engineers should also calculate expected deflection, sensor resolution, natural frequency, fatigue loading, and whether the motor can compensate for spring motion at the required speed. Ignoring gearbox backlash between the encoders can corrupt the inferred torque.

Another mistake is applying τ = kθ outside the spring’s calibrated linear range. Real flexures may show hysteresis, temperature drift, manufacturing variation, or nonlinear stiffness, so experimental torque-deflection calibration is essential. Students should draw the sequence motor–gearbox–spring–load and clearly identify where position is measured.

For exams, distinguish series elasticity from a parallel spring. A series spring lies in the transmitted load path and enables force estimation from deflection; a parallel spring shares load with the motor and often reduces energy consumption. Remember the central compromise: lower stiffness improves compliance and sensing, while higher stiffness improves positional bandwidth.

Conclusion

A series elastic actuator converts controlled compliance into measurable torque, impact tolerance, and more adaptable robot motion. Successful design requires balancing spring stiffness, deflection, bandwidth, fatigue strength, and feedback control rather than treating the spring as an isolated component. Explore more mechanical engineering topics on Mechtics, and share your actuator-design questions in the comments.