B. M. Moores, N. R. Carmichael, K. E. Lee, A. H. Hansen, D. Langlois, H. L. Bartlett, and M. K. Shepherd, "On Ankle-Foot Prostheses: An Introduction to Biomechanical Goals, Modern Design Principles, Device Evaluation, and Current Clinical Practice," submitted to *IEEE Transactions on Neural Systems and Rehabilitation Engineering*, 2026. Manuscript under review.
ABSTRACT: Prosthetic ankle-foot design is a rapidly evolving field encompassing three principal device categories: passive, active, and quasi-passive. Although each category has generated substantial published literature, this body of work remains fragmented, leaving clinicians unaware of emerging research developments and engineers disconnected from clinical practice. This disconnect ultimately impedes the translation of insights between disciplines. Here, we synthesize principles from modern prosthetic ankle-foot design, establishing connections among biomechanical objectives, clinical evaluation methodologies, and engineering design fundamentals across devices used in both clinical practice and research settings.
This review covers passive, active, and quasi-passive ankle-foot prostheses along with the biomechanical and clinical context needed to understand them. My contribution was the section on powered (active) devices, so only that material is summarized here, along with its two figures. The full review, including the passive and quasi-passive sections and the clinical practice content written by my coauthors, will be available through the journal once the paper is accepted.
A) Common high speed (blue) or high torque (purple) motors, paired with transmissions that could achieve the torques and speeds of common activities. B) In this simulation, the motor outputs are driven with realistic human ankle kinematics, and the motors are commanded via open-loop current control to provide a realistic ankle torque for a walking gait. The impact of reflected inertia can be seen in the comparatively poor performance of the small motor with a large transmission. C) The reachable torque/velocity of two realistic motors with different gear reductions, showing that both can emulate the torques and speeds of most biomechanical tasks. D) Strengths and weaknesses of common transmission elements.
Normative gait requires significant energy to be injected into the ankle during fast walking, uphill walking, running, and jumping, and passive prostheses are limited in their ability to emulate these kinetics and kinematics. Powered ankles were developed to address that deficiency, but commercially available devices have seen limited clinical adoption because the benefits of added power largely fail to justify the added noise, weight, maintenance, and cost. Studies to date show mixed results with respect to biomechanical or energetic improvements. As mechatronics technology and control approaches advance, however, the viability of active devices continues to increase.
Almost all powered prostheses rely on electromagnetic motors, which have much higher speed and lower torque than the biological muscles they replace, so a transmission is required and the two are typically co-designed. Brushless DC motors dominate the field for their robustness, low noise, and high torque density. Motor torque is proportional to current and limited primarily by heat, though motors can tolerate brief periods above their nominal current rating, which makes over-currenting common in prosthetic ankles since peak torque is only needed during push off. A key constraint is reflected inertia: rotor inertia felt at the output scales with the square of the transmission ratio, so a small motor behind a 300:1 reduction feels 90,000 times more resistant to acceleration at the joint. Quasi-direct-drive designs use larger diameter motors and much lower ratios to reduce this effect, at the cost of added motor size and mass.
Transmission selection balances mechanical advantage, weight, robustness, efficiency, cost, noise, and packaging. Prismatic elements such as lead, roller, and ball screws are force dense and package well inside the form factor of an ankle, and linkages paired with them can create non-constant mechanical advantage that favors human biomechanics by delivering high torque at specific angles. Rotary-to-rotary elements offer a wider range of ratios but tend to be larger in diameter: gears and belts reach roughly 5:1 with limited peak torque, planetary reducers reach about 9:1 in a compact package, and cycloidal and strain-wave gears reach 30:1 to 100:1 at high torque with added complexity and cost.
Adding a spring in series with the transmission changes the actuator's dynamics and opens up new control paradigms. A stiff actuator generates large force spikes on contact and requires high feedback gains that induce contact instability, while a series elastic actuator deflects its spring in a controlled and measurable way so the loop can be closed around that deflection. The trade-off is reduced large-force bandwidth in exchange for better disturbance rejection, which is usually favorable for prosthetics, though the approach demands extra sensing and springs that are large and heavy. Springs in parallel with the actuator can instead offload actuator torque and allow the motor to be downsized, but their usefulness depends on engagement direction and rest angle, and in some tasks the motor has to fight the spring.
A) A controller that applies a predetermined torque profile through gait phase estimation. B) A finite state machine controller that uses impedance control with parameters tuned for different phases of gait. C) EMG control may be used to convert user intent directly to position or torque.
Prosthetic control is structured hierarchically. Low-level controllers run at roughly 1000 Hz and translate desired positions, velocities, or torques into motor commands. Mid-level controllers run at roughly 100 Hz and convert knowledge of the current task into those commands, commonly through a finite state machine that applies impedance control with parameters tuned per gait phase, or through a continuous phase variable that applies a predetermined torque profile. Finite state machines require reliable transition detection and time-consuming tuning for every state and activity, which has motivated unified phase-based control laws and data-driven parameter optimization. Electromyography can modulate torque or phase directly from residual limb muscle activity, but noise and the absence of proprioceptive feedback make feedforward EMG control alone insufficient for stable walking. High-level controllers handle task classification and environment recognition, increasingly with deep learning, and transitions between activities remain a key challenge since a late or incorrect classification can cause the foot to plantarflex during swing and create a tripping hazard.
