Source-linked AI summary
Design and Validation of a Lightweight, Low-Profile Powered Knee Prosthesis with Quasi-Direct Drive Actuation
Ross J. Cortino, Ryan Posh, Emily G. Keller, Robert D. Gregg
TL;DR
Existing powered knee prostheses must balance positive-work capability against the weight and bulk that hinder QDD clinical translation. This paper designs and validates a lightweight QDD knee using optimized transmission and thermal approaches, demonstrating controlled operation and participant-level performance across walking and sit-stand transitions. The device combines a 2.6 kg mass, 24.5 cm build height, 145 Nm peak torque, low backdrive torque, and reduced acoustic noise, while ramp and stair activities remain to be evaluated.
Problem
Prior QDD knee prototypes provided low-impedance benefits but were too heavy and bulky for clinical translation, despite the need for powered assistance during demanding activities.
Method
The paper designs a lightweight fully-powered knee around an optimized 18:1 two-stage QDD transmission, using thermal modeling and structural analysis, then validates it with benchtop and amputee-participant experiments.
Results
The device demonstrated biomimetic mechanical performance during walking and sit-stand transitions, with three participants, reduced acoustic noise, low backdrive torque, and accurate position and torque control.
Takeaways & Limitations
The results support the clinical feasibility of a lightweight QDD knee that retains low-impedance benefits across knee flexion angles and loading conditions.
Takeaways & Limitations
The study did not evaluate ramp or stair activities with amputee participants, and continuous operation was estimated at under 2 hours for a 102 kg user.
Abstract
from arXiv · showhide
Fully-powered knee prostheses, unlike traditional passive knees, can perform controlled positive work, reducing the need for compensatory behaviors by users during energy-intensive activities. While quasi-direct drive (QDD) actuators provide superior torque control, backdrivability, and acoustic noise properties compared to traditional highly-geared actuators, prior QDD prototypes have been too heavy and bulky for commercial translation. In this work, we present the design and validation of a new lightweight (2.6 kg) and low-profile (24.5 cm tip-to-tip build height) QDD knee prosthesis. By optimizing an 18 to 1 two-stage transmission alongside thermal and structural finite-element analyses, we significantly reduce device mass while enabling a peak torque of 145 Nm. Through benchtop tests, we validate the device's high output torque, low backdrive torque (1 Nm), and its precision position and torque control capabilities. We also demonstrate biomimetic kinematics and peak knee extension torques (within one standard deviation of able-bodied references) during both level-ground walking and sit-stand transitions performed by three participants with transfemoral amputation and varying K-levels. By meeting or improving upon the mass, build height, peak torque, and acoustic noise of a leading commercial powered knee, this work establishes the clinical viability of emerging QDD prostheses that promise improved dynamic performance for their users.
I. INTRODUCTION
Powered knees can provide positive work for demanding activities, but existing designs face weight, size, impedance, and complexity trade-offs. This work introduces a compact QDD knee intended to address those barriers while preserving biomimetic performance and clinical usability.
- Motivation: Millions experience limb loss annually, and conventional quasi-passive knees cannot provide the net-positive work required for ramp ascent, stair ascent, or sit-to-stand transitions.Users instead rely on compensatory behaviors associated with strain on intact joints, increased cognitive load, falls, and asymmetric gait.
- Motivation: Powered knees can perform biomimetic positive and negative work, improving inter-limb symmetry and reducing peak hip flexion moments and intact-joint work during locomotion.These biomechanical benefits are accompanied by increased ipsilateral hip extension moments and intact-side ankle work compared with conventional prostheses.
- Design Challenge: Lightweight powered-knee design remains an engineering challenge because high knee torque demands often require large transmission ratios or added passive elements.Variable transmissions and spring-based architectures can introduce torque degradation at high flexion, increased output impedance, or greater design and manufacturing complexity.
- Related Work: QDD actuators use constant low gear ratios (≤24:1) with high-torque motors to reduce impedance and improve controllability while producing biomimetic torque levels across activities of daily living.Reported benefits include backdrivability, high control bandwidth, accurate torque and impedance control, impact mitigation, and reduced acoustic noise.
- Related Work: Prior QDD prostheses achieved high torque and low backdrive torque but remained heavy and tall, causing alignment problems, excluding shorter users, and increasing prosthetic-side hip effort.A previous QDD leg used a 22:1 transmission, produced 182 Nm peak output torque, and weighed approximately 3.5 kg at the knee.
- Contribution: The presented work introduces a lightweight, compact, fully-powered QDD knee with a 2.60 kg mass, 24.5 cm build height, 18:1 transmission, and 145 Nm peak active torque.The authors also report benchtop dynamic validation, participant evaluations during walking and sit-stand transitions, and reduced acoustic noise relative to the Ossur Power Knee.
II. HARDWARE DESIGN
The knee was designed around an 18:1 QDD transmission to combine biomimetic torque capacity with reduced mass, compact dimensions, and low reflected inertia. Thermal, structural, and multi-objective gear optimization informed the final architecture.
- Design targets: 2.60 kg with batteries and 24.5 cm build height characterize the final knee assembly.The device is 6.3 cm shorter than the Power Knee and has comparable transverse dimensions.
- Actuator performance: The 18:1 actuator has estimated rotor-reflected inertia of 0.037 kg·m² and calculated actuator inertia of 0.041 kg·m².Both values are lower than those reported for the prior QDD design.
- Transmission selection: 18:1 is the minimum transmission ratio selected to contain able-bodied torque-velocity trajectories while limiting temperature, backdrivability, and reflected inertia penalties.The motor must repeatedly reach approximately 8.06 Nm to produce the desired 145 Nm peak joint torque.
- Thermal design: Thermal simulations modeled 600 stair-ascent strides for a 102 kg user, and the motor frame was designed to act as a heat sink for safe temperatures.Across simulations, higher transmission ratios were associated with lower motor winding temperatures.
- Transmission optimization: A genetic-algorithm optimization matched the total 18:1 ratio while minimizing transmission mass and the output-gear radius at the knee center of rotation.The design variables include gear pitch radii, modules, face widths, tooth counts, and helix angles.
- Transmission architecture: 9:1 SPC-PGT and 2:1 secondary gearing form the two-stage transmission, with the secondary stage moving the knee center of rotation proximally.Tempered 420 stainless steel enabled more aggressive material removal and narrower gear face widths than the prior design.
D. Electronics
The electronics use a Raspberry Pi Compute Module 5 and custom interface architecture for high-fidelity control, modular sensing, and simplified communication. Auxiliary sensors support force, contact, center-of-pressure, and thigh-kinematic measurements.
- Control electronics: The Raspberry Pi Compute Module 5 integrates CAN, UART, I2C, and SPI through a custom interface board.CAN communication with the motor driver simplifies the wiring harness.
- Sensing: A distal 6-axis load cell measures ground-reaction forces and moments, foot contact, and center of pressure.A custom amplifier and ADC digitize the load-cell signal for SPI acquisition by the Compute Module 5.
E. Mechanical Structure
The mechanical structure packages the motor, transmission, electronics, wiring, and batteries in lightweight aluminum clamshell housings. Thermal and structural analyses, modular covers, and interchangeable bumpers support durability, cooling, serviceability, and acoustic control.
- Housing: 7075-T6 aluminum clamshell housings hold the motor, transmission, wiring harness, and batteries.Thermally conductive epoxy bonds the motor into the housing to improve heat transfer while keeping the exterior safe to touch.
- Range of motion and noise: Interchangeable TPU bumpers modify range of motion and reduce acoustic noise during ambulation.Without bumpers, the knee ranges from -5° hyperextension to 105° flexion.
- Structural and thermal analysis: Finite-element analysis assessed housing rigidity and thermal performance while targeting loads from a 116 kg user with a factor of safety of 3.The housing was designed to minimize component mass while meeting the structural requirements.
- Packaging and serviceability: The internal cavity provides mounting points and accessible magnet-held covers, while direct frame mounting enables passive electronics cooling.Additional features secure wiring with Velcro or zip ties.
III. EXPERIMENTAL VALIDATION
Validation combined benchtop characterization of actuator capabilities with amputee experiments under walking and sit-stand loading. The tests assessed backdrive, position and torque tracking, peak torque, and function under real-world conditions.
- Validation strategy: Benchtop and in-vivo experiments evaluated intrinsic actuator capabilities and device function during high-velocity and high-torque activities.The in-vivo tasks included level-ground walking and sit-stand transfers.
- Closed-loop position tracking: 33–48 ms estimated and 35–54 ms measured rise times demonstrated a responsive position-control loop for 5°, 10°, and 15° steps.Estimated and externally measured joint angles were compared against commanded positions.
- Open-loop torque tracking: 10.19–17.40 ms measured rise times were obtained for 20 Nm and 40 Nm torque steps.Torque was estimated from motor current and compared with external load-cell measurements.
- Biomimetic torque tracking: 0.05 Nm/kg stair-ascent and 0.07 Nm/kg walking peak-extension discrepancies were smaller than the corresponding able-bodied standard deviations.The reference standard deviations were ±0.17 Nm/kg for stair ascent and ±0.12 Nm/kg for walking.
- Peak torque capacity: 145 Nm was reached in a benchtop peak-torque test using manually applied incremental loads through an extended pylon.A position controller maintained a fixed knee angle during loading.
B. Amputee Participant Experiments
The prosthesis was evaluated with three transfemoral amputee participants during level-ground walking and sit-stand transitions using the HKIC framework. It produced biomimetic walking and transition biomechanics while operating more quietly than the commercial benchmark, although ambient noise may have inflated acoustic measurements.
- Experimental Methods: Three participants—two K4 and one K3—performed level-ground walking at 0.8, 1.0, and 1.2 m/s plus sit-stand transitions.The cohort included participants with varying mobility levels and prior powered-prosthesis experience.
- Experimental Methods: HKIC used unchanged stance-impedance and swing-kinematic models from a previous knee-ankle prosthesis to assess device performance under realistic loading.The presented prosthesis lacked an actuated ankle, so these tests were intended to evaluate device performance rather than the controller itself.
- Level-Ground Walking Performance: Walking peak kinematics and kinetics largely fell within one standard deviation of able-bodied reference data across participants and speeds.The prosthetic biomechanics slightly led the able-bodied reference because shortened prosthetic-side stance shifted swing kinematics earlier in the normalized gait cycle.
- Sit-Stand Transition Performance: Sit-stand transitions produced near-biomimetic joint torques while accommodating individual preferences and timing through the phase-based controller.The thigh-based phase variable allowed participants to indirectly control progression, producing temporal differences across participants and relative to able-bodied references.
- Acoustic Noise Results: 48.54 dB average acoustic noise was approximately 3.9 dB lower than the Power Knee’s 52.33 dB during 1.0 m/s walking.The presented prosthesis also had a 51.65 dB peak versus 57.17 dB for the Power Knee, while approximately 39 dB ambient noise may have inflated measurements.
IV. DISCUSSION
The discussion presents the prosthesis as a lightweight, compact QDD platform with low impedance and biomimetic performance across heterogeneous users. It also identifies a range-of-motion limitation for users who rely heavily on the prosthetic limb during sit-stand transitions.
- A. Device Capabilities Compared to State-of-the-Art: 145 Nm active torque across knee flexion up to 105° distinguishes the design from the Utah knee, whose active torque is limited to 115 Nm and degrades at 88°.The presented design weighs 700 g more than the Utah Direct Ball Screw Drive Knee but avoids its high-flexion transmission singularity.
- A. Device Capabilities Compared to State-of-the-Art: < 1 Nm static backdrive torque and 0.037 kg·m² reflected inertia support accurate impedance and position control without an output torque sensor or encoder.The reflected inertia is 33.9% smaller than the previous QDD design and remains constant across loading profiles.
- B. Amputee Participant Experiments: Comparable knee biomechanics were observed against able-bodied references across three participants spanning K3–K4 mobility levels and diverse anthropometrics.P3’s greater prosthetic-side reliance during sit-stand transitions appeared as increased flexion while sitting, standing hyper-extension, and a longer high-torque regime.
- A. Device Capabilities Compared to State-of-the-Art: 48.54 dB average and 51.65 dB peak noise levels were below the Power Knee under identical control and task conditions.The authors note that A-weighted sound levels do not fully capture perceived loudness or annoyance, although users clearly perceived differences in noise and pitch.
- B. Participant Feedback: A 105° hardstop prematurely blocked P3’s nearly 120° peak knee flexion during sit-stand transitions.The authors identify designing around average able-bodied biomechanics as a limitation for users with reduced intact-limb strength and range of motion.
C. Limitations and Future Work
The device demonstrates commercially relevant QDD performance, while the authors identify remaining limits in mass, endurance, and activity validation. Future work targets broader clinical testing and an integrated knee-ankle platform.
- Mass: 0.20 kg could be saved by replacing the commercial 6-axis load cell with a lighter sensor or alternative force-estimation method.The current 2.6 kg device is heavier than the approximately 1.9 kg Utah Knee.
- Endurance: Under worst-case power draw, the device is estimated to support approximately 8,300 continuous walking strides at 1.0 m/s for a 102 kg user, or under 2 h of operation.LiPo batteries provide high specific power for transient currents but lower specific energy than the lithium-ion cells used in comparator devices.
- Validation scope: Ramp and stair activities have not yet been evaluated with amputee participants, motivating broader ADL testing and adaptation of the HKIC controller for standalone-knee use.The demonstrated evaluations covered benchtop and in-vivo torque control, walking, and sit-stand transitions.
- Contribution: The platform supports tuning-free biomimetic control through high-fidelity position and torque tracking while retaining constant low-ratio QDD behavior across loading conditions.The authors connect these properties with commercially viable mass, reduced mechanical complexity, and quieter operation.
- Future work: Future work includes larger-cohort clinical testing, variable locomotor tasks, and direct comparisons with other fully powered devices.A structurally matching QDD ankle is also intended to enable an integrated knee-ankle platform and system-level comparisons.
a Lightweight, Low-Prole Powered Knee Prosthesis with
The paper concerns quasi-direct drive actuation and presents an electrical-system block diagram alongside transmission gear parameters.
- “Quasi-Direct Drive Actuation” identifies the paper’s actuation approach.
- The electrical-system block diagram shows communication direction and component connections.UART is purple, SPI is indigo, and CAN is teal.
- Table I reports the transmission gear parameters.
II. EXPERIMENTAL VALIDATION
The experimental-validation section includes a table of participant information.
- Table II presents participant information for the experimental validation.
- The supplied passage identifies the participant-information table but does not report individual participant characteristics.
- The table heading establishes participant characteristics as part of the validation report.
A. Motor Characterization
The motor-characterization workflow combines controlled current, velocity, and back-EMF tests with regression-based parameter estimation. These measurements produced calibrated motor parameters for torque modeling and thermal analysis.
- Model: The motor model relates torque to q-axis current, rotor velocity, rotor acceleration, damping, friction, and inertia.The characterization framework defines the motor’s q-axis torque constant, viscous damping coefficient, Coulomb friction magnitude, and rotor inertia.
- Constant Current Test: The constant-current test locks the rotor, commands ±0–10 A, averages five trials, and fits torque constant by linear regression.
- Back-EMF Test: The back-EMF test drives the motor at constant no-load speeds while measuring line-to-line voltage across two winding leads.The driving motor was commanded over velocities from ±0–500 RPM.
- Constant Velocity Test: The constant-velocity test estimates viscous damping and Coulomb friction from steady-state current across velocities from ±0–500 RPM.Five trials were performed at 50 RPM intervals after excluding transient start and end periods.
- Combined Constant Current and Velocity Test: The combined test varies driving velocity and commanded current, averages steady-state measurements, and estimates all three unknown motor parameters.The test spans ±0–500 RPM in 100 RPM increments and ±0–10 A in 1 A increments.
- Results and Outcomes: 0.175 Nm/A was the average measured power-invariant torque constant, exceeding the manufacturer value of 0.169 Nm/A.The measured constant improves the current-to-torque relationship used in the motor’s thermal model.
B. Device Thermal Modeling
The device’s thermal modeling combines experiments, electrical-domain thermal circuits, and CAD-based simulation to predict winding and frame temperatures under load. Validation showed predictions within the thermal camera’s documented error, supporting thermally informed actuator design.
- Motivation: Joule heating limits motor performance because higher current raises winding temperatures, risking demagnetization and motor failure.The model is used to select transmission and frame designs that keep temperatures acceptable while reducing motor load.
- Motor Thermal Model: A motor thermal model represents winding heat transfer through conduction and convection, using winding temperature, ambient temperature, thermal resistance, thermal capacitance, and Joule-heating heat flux.The parameter-identification experiment applied 9 A to two windings and monitored temperatures and voltage until steady state.
- Frame Thermal Model: A frameless, potted motor requires a frame in direct contact with the epoxied windings, so the frame model includes conduction between the windings and frame.The updated electrical-domain model adds frame thermal coefficients and represents heat transfer among windings, frame, and ambient air.
- Frame Thermal Model: The frame model defines thermal capacitance and resistances for winding-to-frame, frame-to-ambient, and winding-to-ambient heat transfer.The winding-to-frame resistance was estimated using a dynamic Ansys Mechanical simulation of the CAD frame model.
- Results and Outcomes: ±2°C was the maximum steady-state discrepancy between predicted and measured motor and frame temperatures, matching the thermal camera’s documented error.Experiments and Ansys Mechanical simulations used a 9 A load with a stock heat sink attached to the motor.
- Results and Outcomes: Validated thermal predictions enable estimation of winding and frame temperatures under biomimetic knee loads, informing lighter transmission and actuator designs.Improved motor thermal performance can support a smaller mechanical advantage and increased backdrivability.