SCIENCE ROBOTICS | RESEARCH ARTICLE

PROSTHETICS Copyright © 2022

The Authors, some
A lightweight robotic leg prosthesis replicating the rights reserved;
exclusive licensee
biomechanics of the knee, ankle, and toe joint American Association
for the Advancement
of Science. No claim
Minh Tran†, Lukas Gabert†, Sarah Hood, Tommaso Lenzi* to original U.S.
Government Works
Robotic leg prostheses promise to improve the mobility and quality of life of millions of individuals with lower-
limb amputations by imitating the biomechanics of the missing biological leg. Unfortunately, existing powered
prostheses are much heavier and bigger and have shorter battery life than conventional passive prostheses,
severely limiting their clinical viability and utility in the daily life of amputees. Here, we present a robotic leg
prosthesis that replicates the key biomechanical functions of the biological knee, ankle, and toe in the sagittal
plane while matching the weight, size, and battery life of conventional microprocessor-controlled prostheses.
The powered knee joint uses a unique torque-sensitive mechanism combining the benefits of elastic actuators
with that of variable transmissions. A single actuator powers the ankle and toe joints through a compliant,
underactuated mechanism. Because the biological toe dissipates energy while the biological ankle injects
energy into the gait cycle, this underactuated system regenerates substantial mechanical energy and replicates
the key biomechanical functions of the ankle/foot complex during walking. A compact prosthesis frame enclos-
es all mechanical and electrical components for increased robustness and efficiency. Preclinical tests with three
individuals with above-knee amputation show that the proposed robotic leg prosthesis allows for common am-

Downloaded from https://www.science.org on March 31, 2024

bulation activities with close to normative kinematics and kinetics. Using an optional passive mode, users can
walk on level ground indefinitely without charging the battery, which has not been shown with any other
powered or microprocessor-controlled prostheses. A prosthesis with these characteristics has the potential to
improve real-world mobility in individuals with above-knee amputation.

INTRODUCTION ambulate in the community (8). Improvements in prosthetic tech-

To date, most prostheses available to individuals with above-knee nologies are necessary to address the unmet needs of the millions of
amputation are passive devices that cannot replicate key biome- individuals living with lower-limb amputation (9).
chanical functions of the missing biological leg. Most ankle/foot Robotic leg prostheses promise to improve the ambulation
prostheses consist of a carbon fiber plate enclosed in a rubber ability of individuals with lower-limb amputation by imitating key
foot shell (1). Some ankle-foot prostheses have an actual ankle biomechanical functions of the missing biological leg with onboard
joint actuated by passive elements such as springs and dampers. actuation systems, sensors, and power supplies (10). Researchers
Virtually all available ankle/foot prostheses rely on the flexibility have proposed different advanced actuation designs to efficiently
of the rubber foot shell to emulate the movement of the metatarsal provide the wide ranges of speed and torques needed to imitate
joint (toe joint), and only one ankle-foot prosthesis available on the the biological leg (6, 7). Powered knee prostheses can be actuated
market has an articulated, passive toe joint (2). For above-knee pa- using springs in series (11) and parallel to a motor (12), multijoint
tients, the ankle/foot prosthesis is connected to a prosthetic knee, actuators (13), antagonistic actuators (14), or high–torque density
which may have a single-joint or polycentric design, passively actu- motors (15). Similarly, powered ankle prostheses can be actuated
ated by springs and dampers (3). In microprocessor-controlled using four-bar mechanisms and polycentric designs (16–18),
prostheses, the mechanical impedance of the ankle and knee joint using a spring in parallel or in series to the motor (19), inductive
can be actively adjusted during gait to facilitate walking at a variable charging (20), or implementing clutches and brakes in combination
cadence while improving stability and reducing the risk of falls (4). with motors (21, 22). These advanced actuation systems have
However, they cannot actively generate movements or inject net- enabled two powered prostheses to reach the market. The Ottobock
positive energy into the gait cycle (5), which are key biomechanical Empower ankle/foot prosthesis uses a series/parallel elastic actuator
functions of biological legs (6, 7). Prosthesis users compensate for (23), whereas the Ossur Power Knee uses a clutchable series-elastic
these deficiencies with their residual limb and contralateral leg, re- actuator (24). Unfortunately, after several years on the market,
sulting in a slower, less efficient, and less stable gait compared with powered prostheses have failed to achieve clinical success (25–28).
nonamputees (8). Because these passive prostheses cannot actively Although this negative outcome likely results from a combination of
generate knee torque, climbing stairs and ramps or transitioning factors, there are key design limitations that affect function and us-
between sitting and standing is much more challenging for individ- ability of existing powered prostheses. These powered devices are
uals with above-knee amputations than nonamputees. As a result, much heavier and bigger and have shorter battery life than their
most individuals with an above-knee amputation are not able to passive counterparts. Increasing the prosthesis weight negatively
affects both biomechanics and clinical outcomes. During walking,
larger prosthesis weight has been correlated to increased metabolic
Department of Mechanical Engineering and Robotics Center, University of Utah, energy cost (29), stance-time and swing-time asymmetries (30), hip
Salt Lake City, UT, USA.
*Corresponding author. Email: t.lenzi@utah.edu
effort (31), and reduced socket stability. Increasing the prosthesis
†These authors contributed equally to this work. build height and the distance between the knee center of rotation

Tran et al., Sci. Robot. 7, eabo3996 (2022) 23 November 2022 1 of 17

SCIENCE ROBOTICS | RESEARCH ARTICLE

and the top of the pyramid has limited the number of people that Biomechanical studies of nonamputee gait suggest that the
can be fitted to the prosthesis. Moreover, powered prostheses have metatarsal (toe) joint plays an important function during gait
much shorter battery life than their passive counterparts—a few (46). Individuals with metatarsophalangeal arthrodesis (fusion of
hours versus a few days—which has a negative effect on usability the toe joint) have decreased step length and reduced plantarflexion
in real life. We need to decrease the weight and size and extend moment in the affected side (47). Studies with a passive prosthesis
the battery life of robotic leg prostheses to improve their clini- emulator show that the stiffness of the toe joint has a substantial
cal effect. effect on ankle power, center of mass, and push-off work during
An alternative design strategy for robotic leg prostheses consists walking (48). Moreover, a previous study has shown improvements
of powering only a subset of activities (32–35), avoiding net-positive in the metabolic cost of walking of individuals with below-knee am-
energy injection, or adjusting the mechanical behavior of the pros- putations using a passive prosthesis with midfoot and metatarso-
thetic joint without actively controlling movements or injecting phalangeal joints (49). Despite this evidence, only one research
energy (36–39). By relaxing the actuation speed and torque require- robotic ankle/foot prosthesis has been designed with a powered
ments, these semi-active and quasi-passive prostheses can be made toe joint (50, 51). Unfortunately, the added toe function comes at
lighter and smaller and still achieve longer battery life than fully the cost of a substantial increase in prosthesis weight due to the ad-
powered prostheses. For example, designing the knee actuator to dition of a dedicated toe actuator. Thus, there is a need for designs
power only stair ambulation (33, 34) or only the swing phase of that can replicate the biomechanical function of the toe joint in a
gait (32, 40) leads to a lighter and smaller prosthesis. Similarly, lightweight prosthesis.
lighter and smaller ankle/foot prostheses can be developed by avoid- Here, we present a robotic leg prosthesis designed to replicate the
ing net-positive energy injection (41), limiting the active control of key biomechanical functions of the biological knee, ankle, and toe
movements to non–weight-bearing activities (38, 42), or adjusting joints in the sagittal plane while matching the weight, size, and

Downloaded from https://www.science.org on March 31, 2024

the mechanical stiffness of the prosthetic joint (37). More recently, battery life of microprocessor-controlled prostheses. As we will
switching gears between different ambulation activities has been show, this level of lightness, performance, and efficiency is
proposed to bridge the gap between semi-active and fully enabled by bioinspired actuation designs combined with a holistic
powered prostheses. However, this design solution comes at the design approach (52). Using biomechanical analysis, analytical
cost of functionality, because the prosthesis cannot provide torque models, and dynamic simulations, we will show that a unique
while switching gears (43–45). Thus, semi-active and quasi-passive torque-sensitive actuator can replicate key biomechanical features
prostheses are typically lighter and smaller and have longer battery of the biological knee joint by combining the benefits of series-
life than fully powered devices but cannot replicate key biomechan- elastic actuators with that of variable transmissions. We will also
ical functions of the missing biological leg. show that an actuator based on a compliant, underactuated mech-
anism can replicate the ankle and toe joint functions while

Fig. 1. System overview. (A) A photo of the Utah Bionic Leg. (B) Partially sectioned view of the prosthesis model highlighting the main electrical and mechanical
components.

Tran et al., Sci. Robot. 7, eabo3996 (2022) 23 November 2022 2 of 17

SCIENCE ROBOTICS | RESEARCH ARTICLE

regenerating substantial mechanical energy during walking. Last, as a variable transmission to change the torque ratio passively, con-
we will also show the ability to walk on level ground indefinitely tinuously, and quickly in response to varying knee extension torque
without charging the battery. A prosthesis with these characteristics by following a specific curve defined by the design geometry. The
has the potential to improve real-world mobility in individuals with ankle/foot module uses a compliant, underactuated mechanism to
above-knee amputations. power both the ankle and the toe joint. This compliant, underactu-
ated mechanism transfers mechanical energy from the toe to the
ankle joint, improving efficiency, while storing and releasing
RESULTS energy in a lightweight spring assembly. The knee and the ankle/
System overview foot modules have independent onboard power supplies and em-
The proposed powered prosthesis—namely, Utah Bionic Leg—con- bedded electronic systems (Fig. 1B). Both modules were designed
sists of independent knee and ankle/foot modules (Fig. 1A). The on the basis of the International Organization for Standardization
knee module uses a unique torque-sensitive actuator that works (ISO) standards, facilitating their use outside the laboratory. The

Downloaded from https://www.science.org on March 31, 2024

Fig. 2. Analysis of the torque-sensitive actuator and the compliant, underactuated mechanism. (A) Biological knee torque and speed during walking and stair
ascent for nonamputee individuals. (B) Predicted torque ratio during walking and stair ascent. (C) Predicted current root mean square (RMS) and maximum winding
voltage for walking and stair ascent. (D) Torque, speed, and power of the ankle and toe during walking and stair ascent. (E) Predicted energy flow analysis for the under-
actuated ankle/toe mechanism and ankle-only design. (F) Predicted electrical energy, actuator peak power, and velocity at peak torque for the underactuated ankle/toe
mechanism and ankle-only design.

Tran et al., Sci. Robot. 7, eabo3996 (2022) 23 November 2022 3 of 17

SCIENCE ROBOTICS | RESEARCH ARTICLE

powered knee module of the Utah Bionic Leg has similar weight and current limit during stair ascent primarily because of the high
size to the Ottobock C-leg—a microprocessor-controlled knee torque required in stance (Fig. 2C). In contrast, the torque-sensitive
prosthesis (53). The powered ankle/foot module has similar actuator can satisfy both motor torque and speed requirements by
weight and size to the Ottobock Meridium—a microprocessor-con- adapting the torque ratio during ambulation. In stance, when the
trolled ankle/foot prosthesis (54). Our powered knee module is knee torque is high and the speed is low, the torque ratio quickly
slightly lighter (40 g, 2.3%) and has a shorter pyramid-to-knee increases from its minimum to its maximum, lowering the required
joint length than the C-Leg Genium (3 mm, 11.5%). Our ankle motor torque for both walking and stair ascent (Fig. 2B). In the
module is slightly taller (4 mm, 5%) and heavier (70 g, 4%) than swing phase, when the knee torque is low and the speed is high,
the Ottobock Meridium, although our powered device includes a the torque ratio stays close to its minimum, reducing the required
custom force/torque sensor (55), which is not present in the Meri- motor speed as well as the inertial torque (Fig. 2B). Because the knee
dium. Our powered knee and ankle/foot prosthesis is substantially speed is low when the torque is high (Fig. 2A), increasing the torque
smaller and lighter than the Ossur Power Knee (24) and the Otto- ratio proportionally to the knee extension torque does not cause the
bock Empower (23)—the only powered devices on the market. Sim- motor to reach the winding limit, although the motor velocity is
ilarly, powered prostheses designed by research laboratories that higher than it would be without the torque-sensitive actuator
have not yet been marketed are between ~50 and ~100% heavier during stance (fig. S4). Thus, the simulations show that by leverag-
than the Utah Bionic Leg, which has an additional powered toe ing the nonsimultaneous peaks of torque and speed, the torque-sen-
joint and embeds all the mechanical and electrical components sitive actuator can reduce the motor speed/torque requirements,
(11, 56–58). Both commercially available and research-powered enabling a small, lightweight motor to efficiently provide the wide
prostheses have a larger joint axis to pyramid length compared ranges of torques and speeds required for a knee prosthesis.
with the Utah Bionic Leg. Although battery life depends on many

Downloaded from https://www.science.org on March 31, 2024

factors, the Utah Bionic Leg can be operated in passive mode, en- Analysis of the compliant, underactuated ankle-toe
abling the user to walk without needing to charge the battery. In this mechanism
passive mode, the knee joint behaves primarily as a damper, Our compliant, underactuated mechanism is inspired by the biome-
whereas the ankle joint behaves primarily as a spring, and they chanical analysis of the biological ankle/foot complex. This analysis
imitate the mechanical behavior of microprocessor-controlled (Fig. 2D) shows that during walking, the toe torque is nearly pro-
knee and ankle/foot prosthesis. This functionality has not been portional to the ankle torque for a large part of the stance phase (20
shown by existing powered or microprocessor-controlled to 60% stride), although the peak torque is much lower for the toe
prostheses. than for the ankle (0.12 versus 1.35 Nm/kg). Moreover, the veloci-
ties of the two joints are comparable in magnitude and opposite in
Analysis of the torque-sensitive knee actuator direction, peaking at −218°/s and 280°/s for the ankle and toe joint,
Our torque-sensitive actuator is inspired by the biomechanical anal- respectively. As a result, the toe dissipates power, whereas the ankle
ysis of the biological knee joint. This analysis shows that the knee generates power. Thus, during walking, the ankle and toe torque are
generates almost four times the torque in extension than flexion nearly proportional, and the combined ankle and toe power is
(1.21 versus 0.3 Nm/kg), and the peak extension torque is more smaller than the power of the ankle alone. This analysis suggests
than double in stair ascent compared with level-ground walking that a single actuator could power both the ankle and toe joint, re-
(1.21 versus 0.51 Nm/kg). In contrast, the knee extension velocity quiring fewer mechanical and electrical components than using two
is three times higher during walking than stair ascent (339°/s separate actuators. Moreover, this analysis shows that the combined
versus 99°/s), although the peaks of the knee flexion velocity are ankle and toe power is smaller than the power of the ankle alone.
similar between the two activities. The stance phase (foot on the Powering both the ankle and toe joints in a prosthesis is challenging
ground) requires much greater torque and lower velocity than the because of the stringent weight and size requirements. Our compli-
swing phase (foot off the ground) for both level-ground walking ant, underactuated mechanism aims to address this issue by en-
and stair ascent (Fig. 2A). Thus, the biological knee joint produces abling a single actuator to efficiently power both the ankle and
wide ranges of torques and speeds during ambulation, but the peak the toe joints.
of knee torque and speed are not simultaneous. Providing wide Dynamic simulations show the function of the compliant,
ranges of torques and speeds with a small and lightweight electrical underactuated system during walking by comparing its perfor-
actuator is challenging, because mechanical power output and elec- mance to that of an equivalent actuator powering the ankle joint
trical efficiency decrease sharply outside narrow ranges of torque only. The energy flow analysis (Fig. 2E) shows that the ankle-only
and speed (59, 60). Our torque-sensitive actuator aims to address design requires 14.4 J per stride of electrical energy to produce 6.3 J
this issue by altering the torque ratio in response to knee extension per stride of mechanical energy at the ankle, achieving an overall
torque. efficiency of 43.8%. In comparison, the underactuated design re-
Dynamic simulations show how the passively variable torque quires 8.2 J per stride of electrical energy, achieving an overall effi-
ratio affects the motor speed/torque requirements by comparing ciency of 76.8%. Thus, the electrical energy consumption per stride
the performance of a torque-sensitive actuator to that of the same is 43.0% lower in the underactuated design (Fig. 2F). The energy
actuator fixed at the maximum or minimum of the torque ratio flow analysis also shows that the reduced electrical energy consump-
range (Fig. 2B). These simulations suggest that using the actuator tion is primarily due to the toe regenerating 4.5 J per stride of me-
at a fixed high-torque ratio would not satisfy the winding voltage chanical energy. Last, the underactuated system shows lower energy
limit primarily because of the high velocity required in the swing losses than the ankle-only design (−1 J per stride of Joule heating
phase (Fig. 2C). Moreover, the simulations show that using the ac- and −0.5 J per stride of friction). These lower energy losses are due
tuator at a fixed low-torque ratio would exceed the maximum to the lower velocity and acceleration of the linear actuator in the

Tran et al., Sci. Robot. 7, eabo3996 (2022) 23 November 2022 4 of 17

SCIENCE ROBOTICS | RESEARCH ARTICLE

underactuated design, which result in lower inertial torque and me- module were essential to allow for passive swing motion (movie
chanical power output at the motor (Fig. 2E). Thus, the simulations S3). In addition, based on the reading of an external six-axis load
show that by leveraging the concurrent torque generation at the bi- cell, the knee torque step response showed steady-state errors
ological toe and ankle joints, a compliant, underactuated design can lower than 1.2% and a rise time of ~32 ms irrespective of torque
enable a single actuator to power both the ankle and toe joint and direction, resulting in a −3-dB bandwidth of ~11 Hz (Fig. 3C),
also reduce electrical energy consumption. which was greater than that of the biological knee (61). Because
the torque-sensitive joint acted in extension only and the torque
Performance verification of the powered knee and ankle/ step response was nearly identical in extension and flexion, we con-
foot modules on the bench clude that the torque-sensitive joint did not affect the bandwidth of
The knee closed-loop step position tests (Fig. 3A) showed rise times the open-loop torque controller (Fig. 3C). The motor current
between 37 and 56 ms, depending on the position step size. Thus, during the torque step response test was 18 to 26% lower in exten-
the −3-dB bandwidth of the knee position controller was between sion than in flexion, because the torque-sensitive joint only changed
9.5 and 6.2 Hz, which exceeded the position bandwidth of the bio- the torque ratio in response to extension torque (Fig. 3C). The con-
logical knee (61). Output impedance tests showed that the tinuous motor current increased from 4.21 A (the rated value) to
minimum backdriving torque measured by the external six-axis 9.91 A (a 135% increase), because the contact between the motor
load cell with the motors off was 0.3 Nm (Fig. 3B), which was and the frame provided enhanced thermal dissipation (Fig. 3D)
lower than other powered knee prostheses [3.2 Nm (58), 2.6 Nm (35). This enhanced thermal dissipation was critical for the
(62), and 3 to 5 Nm (56)]. On the basis of the system identification powered knee module to provide biomechanically appropriate
of the output impedance (Fig. 3B), the damping and inertia at the function without overheating (see Supplementary Methods).
output joint were 0.43 Nm/rad and 0.04 kg·m2, respectively. The low Locking the torque-sensitive joint removed the ability of the knee

Downloaded from https://www.science.org on March 31, 2024

minimum backdriving torque and output impedance of the knee module to increase the torque ratio (Fig. 3E). With the torque-

Fig. 3. Characterization of the knee module on the bench. (A) Step response of the knee closed-loop position controller. (B) Knee backdriving torque and estimated
impedance. (C) Knee torque step response and corresponding motor current. (D) Temperature of the knee motor housing/frame for different motor currents and cor-
responding knee frame thermal imaging. (E) Bell-shaped knee torque input and corresponding knee torque ratio and knee Joule heating losses. (F) Tracking of repre-
sentative position profiles for the swing phase of walking and stair ascent.

Tran et al., Sci. Robot. 7, eabo3996 (2022) 23 November 2022 5 of 17

SCIENCE ROBOTICS | RESEARCH ARTICLE

sensitive joint locked at the bottom end, the Joule heating losses in- with the ankle joint, the peak velocity of the linear actuator reduced
creased by 140%; thus, the knee could not provide more than 90 Nm from ~25 to 1 mm/s, a 95% decrease (Fig. 4B). Thus, the underac-
without overheating. This torque level was 40% lower than the 150- tuated mechanism could substantially reduce the required motor
Nm limit with the torque-sensitive mechanism unlocked and is not speed when the toe and ankle joint moved in phase. The ankle
sufficient for most people in the United States (75% of adult males torque step response (Fig. 4C) showed rise times between 32 and
and 46% of adult females) (63) to naturally climb stairs or to stand 25 ms, depending on the torque direction. These differences were
up from a seated position. Locking the torque-sensitive joint at the consistent with the measured stiffness of the series spring
top end can address this torque limitation. However, it lowered the (Fig. 4D) in tension (829 N/mm) and compression (1273 N/mm).
maximum knee joint velocity substantially due to voltage saturation For the tested conditions, the bandwidth of the ankle torque con-
at the motor, and the knee module could not perform natural swing troller was between 14 and 11 Hz, which exceeded the torque gen-
phase movements in walking and stair ascent (Fig. 3F), which may eration bandwidth of the biological knee and ankle (61). The
cause the user to scuff, stumble, and fall. Thus, the torque-sensitive continuous current increased from 7.58 A (the rated value) to
joint was essential for the knee module to satisfy basic torque and 10.30 A, a 36% increase (Fig. 4E), because the motor was in
speed requirements for ambulation. contact with the frame. Thus, the thermal dissipation was necessary
The ankle closed-loop step position tests (Fig. 4A) showed rise for the ankle module to provide biomechanically appropriate func-
times between 30 and 61 ms, resulting in −3-dB bandwidths tion without overheating (see Supplementary Methods). Output
between 12 and 6 Hz. The sine-wave position test of the underactu- impedance tests (Fig. 4F) with an external six-axis load cell
ated ankle mechanism showed that if the toe joint moved in phase showed that the minimum backdriving torque with the motors

Downloaded from https://www.science.org on March 31, 2024

Fig. 4. Characterization of the ankle module on the bench. (A) Step response of the ankle closed-loop position controller. (B) Linear actuator speed for sinusoidal
position input at the ankle and fixed toe and for sinusoidal position input at both the ankle and the toe. (C) Ankle torque step response. (D) Linear spring characterization
in compression and extension. (E) Temperature of the ankle motor housing/frame for different motor currents and corresponding ankle frame thermal imaging. (F) Ankle
backdriving torque and estimated impedance.

Tran et al., Sci. Robot. 7, eabo3996 (2022) 23 November 2022 6 of 17

SCIENCE ROBOTICS | RESEARCH ARTICLE

off was 1.2 Nm, which was about half of other powered prostheses passive prosthesis. Both the maximum ankle plantarflexion torque
tested under similar conditions [3.2 Nm (58) and 3 to 5 Nm (56)]. and dorsiflexion angle of the ankle were within 10.0% of the passive
System identification estimated the damping and inertia at the prosthesis reference. In the swing phase, the ankle angle is near 0°,
output joint to be 2.18 Nm/rad and 0.35 kg·m2, respectively. and little active dorsiflexion movement is shown. In addition, in the
swing phase, the knee achieved 63.8° ± 1.3°, showing an extension
Validation of the powered knee-ankle-toe prosthesis timing within 0.7 ± 0.2% of passive prosthesis reference, which
during common ambulation tasks allowed for proper foot clearance. Thus, the Utah Bionic Leg
Clinically relevant ambulation goals derived from nonamputee bio- enabled the amputee participants to walk safely in both standard
mechanics (64–67) provided a reference to assess the performance mode and passive mode, although the standard mode better satis-
of the Utah Bionic Leg during ambulation with three above-knee fied the clinically relevant ambulation goals. Moreover, locking the
amputee participants (Fig. 5). Table S6 provides an in-depth assess- toe reduced electrical efficiency and mechanical power output of the
ment of the performance, and main results are summarized ankle module.
hereafter. When climbing stairs one step at a time (Fig. 6), the powered
During walking in standard mode, the knee joint was slightly knee generated biomechanically accurate torque, peaking at
flexed at heel strike (9.3° ± 3.1°, mean ± SD), and stance knee 1.08 ± 0.08 Nm/kg, within 1.4 ± 1.4% of able-bodied timing.
flexion peaked at 13.7° ± 5.0°. However, there were substantial dif- However, the prosthetic knee joint extended faster than the biolog-
ferences between participants. For participants 1 and 3, the stance ical knee, which resulted in a peak mechanical power 28.5% higher
knee flexion was minimal. Thus, the knee extension torque was gen- than the physiological value (3.25 ± 0.19 W/kg). In contrast, the
erally low, and the torque-sensitive joint stayed close to its maximum ankle plantarflexion torque (0.62 ± 0.07 Nm/kg) was
minimum. In contrast, participant 2 showed physiological stance considerably lower than the normative value (1.27 Nm/kg). When

Downloaded from https://www.science.org on March 31, 2024

knee flexion and knee extension torque. The torque-sensitive climbing stairs two steps at a time, the knee provided up to
joint extended as expected from simulations, causing the knee 1.67 ± 0.13 Nm/kg of extension torque and 4.71 ± 0.36 W/kg of pos-
torque ratio to increase proportionally. In addition, in standard itive mechanical power, which corresponded to a 54.6 and 44.9%
mode, the ankle plantarflexion torque at push-off reached increase, respectively, compared with climbing stairs one step at a
1.50 ± 0.10 Nm/kg, which was within 2% of normative value and time. In the swing phase, the knee and the ankle module closely fol-
within 5% of normative timing. The maximum angle and torque lowed the able-bodied reference (table S6), safely clearing the step
of the toe joint were 36.3° ± 1.1° and 0.12 ± 0.02 Nm/kg, respective- and positioning the prosthetic foot in preparation for the next step
ly, which were close to biomechanical goals of 39.6° and 0.12 Nm/ to be taken. In the stance phase, the knee torque ratio increased pro-
kg. In the swing phase, both the knee and the ankle joint trajectory portionally to the knee extension torque until it reached its
closely followed the normative data. Walking in standard mode with maximum for all stair climbing conditions. In the swing phase,
the toe joint locked had a visible effect on the ankle kinetics and the knee extension torque is low; hence, the torque ratio stayed
kinematics (fig. S10). The peak motor velocity increased by 47%, close to its minimum (Fig. 6). Thus, the Utah Bionic Leg provided
resulting in a 56% increase in the peak motor mechanical power. appropriate torque, clearance, and foot placement to enable climb-
Because of the motor approaching the winding limitation, the ing stairs both one and two steps at a time, which was not possible
ankle mechanical energy decreased by ~18% from 0.17 J/kg with with microprocessor-controlled prostheses.
the toe unlocked to 0.14 J/kg with the toe locked. Despite the In stair descent, the powered knee generated up to 1.05 ± 0.12
lower ankle mechanical energy, the electrical energy consumption Nm/kg of extension torque, which was within 25% of nonamputee
increased by 24% with the toe locked. In passive mode, there was no data (1.34 Nm/kg). The ankle dorsiflexed up to 19.3° ± 0.7°, slowing
active ankle push-off or visible stance knee flexion, similar to a down the stair descent movement. In addition, the toe showed a

Fig. 5. Above-knee amputee participants ambulating with the Utah Bionic Leg. Representative images of the three participants ambulating on the treadmill and
staircase.

Tran et al., Sci. Robot. 7, eabo3996 (2022) 23 November 2022 7 of 17

SCIENCE ROBOTICS | RESEARCH ARTICLE

Downloaded from https://www.science.org on March 31, 2024

Fig. 6. Kinematics and kinetics across five ambulation activities, with joint angles, joint moments, and knee torque ratio recorded with three amputee par-
ticipants. Colored lines denote mean values across 10 strides. Shaded areas in gray denote nonamputee biomechanical reference, and shaded areas in purple denote
passive prosthesis reference. For knee torque ratio plots, shaded areas in blue denote the bounds of the torque ratio variation.

Tran et al., Sci. Robot. 7, eabo3996 (2022) 23 November 2022 8 of 17

SCIENCE ROBOTICS | RESEARCH ARTICLE

peak extension of 36.0° ± 1.0°, enabling the prosthesis to maintain of mechanical energy per stride, respectively. Subtracting the
stable contact with the step. In the swing phase, the prosthetic knee energy losses due to Joule heating and friction, the knee motor re-
flexed to safely clear the step, with timing similar to able-bodied quired 0.89 ± 0.10 and 2.52 ± 0.29 J/kg of electrical energy to climb
data (1.5 ± 1.7%). In addition, in the swing phase, the ankle plantar- one and two steps at a time, respectively. In contrast, the ankle
flexed up to 9.9° ± 0.2°, enabling the foot to contact the subsequent motor regenerated small amounts of electrical energy (table S7).
step with the toe first. Consistently with the controller in use (68), Thus, the combined knee and ankle/foot prosthesis requires
there were visible differences between the ankle kinetics and the 0.88 ± 0.17 and 2.55 ± 0.29 J/kg of electrical energy per stride for
nonamputee reference. Similar to stair ascent, the knee torque stair ascent and double stair ascent, respectively. In stair descent,
ratio reached its maximum during stance when the knee extension the knee motor absorbed 1.06 ± 0.05 J/kg of mechanical energy
torque was high, and it stayed close to its minimum in swing per stride, regenerating 0.17 ± 0.20 J/kg of electrical energy. The
(Fig. 6). The Utah Bionic Leg provided sufficient torque and ankle motor absorbed 0.12 ± 0.09 J/kg of mechanical energy per
range of motion to enable descending stairs while placing the stride and regenerated 0.07 ± 0.05 J/kg of energy. Thus, in stair
whole foot on the step. descent, the combined knee and ankle/foot prosthesis regenerated
0.24 ± 0.17 J/kg of electrical energy per stride.
Actuation performance and battery life
The actuation performance during ambulation was assessed by es-
timating mechanical and electrical energy consumption based on DISCUSSION
the actuator models and the embedded sensors (see Supplementary Providing wide ranges of torque and speed is necessary for powered
Methods). These estimates showed that, when walking in standard knee prostheses to replicate the biological knee biomechanics
mode, the knee joint absorbed 0.27 ± 0.04 J/kg of mechanical energy during ambulation (Fig. 2A). Satisfying these requirements with a

Downloaded from https://www.science.org on March 31, 2024

per stride, whereas the ankle joint generated 0.23 ± 0.03 J/kg and the lightweight and compact actuator is challenging, because efficiency
toe absorbed 0.04 ± 0.01 J/kg. After accounting for Joule heating and mechanical power output of electrical motors decrease sharply
and friction losses (table S7), the knee motor was estimated to re- outside of a small operating range. Our analytical models show that
generate 0.03 ± 0.06 J/kg of electrical energy, whereas the ankle/toe a torque-sensitive actuator can address this problem by working like
motor was estimated to consume 0.35 ± 0.05 J/kg of electrical a passive variable transmission (see Supplementary Methods).
energy. Thus, the total electrical energy consumption estimate in Benchtop experiments confirm the model predictions by showing
standard mode was 0.32 ± 0.03 J/kg per stride. When walking in that when increasing the torque ratio proportionally to the knee ex-
passive mode, the knee, ankle, and toe joints absorbed tension torque, the required motor torque and current were sub-
0.15 ± 0.01, 0.03 ± 0.02, and 0.03 ± 0.01 J/kg, respectively, of me- stantially reduced (Fig. 3C). The motor torque decreased
chanical energy per stride. After accounting for Joule heating and proportionally to the torque ratio, but the Joule heating losses de-
friction losses (table S7), the knee motor was estimated to regenerate creased quadratically (Fig. 3E), which enabled a small motor (22-
0.11 ± 0.01 J/kg of electrical energy, and the ankle/toe motor was mm diameter, 170 g) to efficiently provide large torque at the
estimated to expend 0.04 ± 0.01 J/kg of electrical energy per knee joint. Although the required motor speed tends to increase
stride. Thus, the combined knee and ankle/foot prosthesis was esti- as the motor torque decreases (fig. S4), the winding voltage limit
mated to regenerate 0.07 ± 0.02 J/kg of electrical energy per stride. is not violated because, during ambulation, the knee velocity is gen-
On the basis of these estimates and accounting for the ~2-W erally low when the knee torque is high (Fig. 2A). The torque-sen-
electrical power consumption of the embedded electronic system, sitive actuator can generate physiological swing trajectories, which
in standard mode, the proposed powered prosthesis would was not possible when the torque-sensitive joint was locked at the
provide up to 7730 ± 905 strides and 15,460 steps with the top end (Fig. 3F). Thus, without the torque-sensitive joint, amputee
current battery (2400 mAh). This battery life exceeded the 10,000 users would not be able to safely ambulate with the proposed knee
(69) and 1500 (70) steps that nonamputee and amputee partici- (Fig. 3). The knee torque bandwidth measured experimentally was
pants, respectively, take on average in a day. Moreover, in passive the same in flexion (nonsensitive to torque) and extension (sensitive
mode, the battery was estimated to recharge at the rate of to torque) (Fig. 3C), confirming that, different from series-elastic
2.0 ± 0.6 J per stride. Additional experiments with a single actuators (71), the compliance of the torque-sensitive actuator did
amputee participant confirmed the estimates of electrical energy not have a negative effect on the torque-control bandwidth. By com-
by directly measuring the battery voltage and current in standard bining a small, lightweight motor (22-mm diameter, 170 g) with a
and passive mode (fig. S11). Specifically, the electrical energy con- relatively low transmission ratio, the torque-sensitive actuator
sumption directly measured at the battery was 0.27 J/kg per stride in achieved low backdriving torque (0.3 Nm) and reflected inertia
standard mode, 22% smaller than the estimated 0.35 J/kg value. In (0.04 kg·m2) (Fig. 3B). The low output impedance of the knee
passive mode, the electrical energy regeneration measured at the joint was essential for achieving electrical energy regeneration in
battery was 0.1 J/kg per stride (fig. S12). As expected, the battery walking. These performance improvements were obtained with a
voltage decreased in standard mode and increased in passive small and lightweight mechanism (~40 g) that fits into an extremely
mode (fig. S13). Thus, the Utah Bionic Leg could regenerate elec- compact knee prosthesis (70 mm max width, 255 mm build height,
trical energy during walking while providing similar kinematics and and 23 mm joint-pyramid distance; Fig. 1). Thus, this study showed
kinetics to a microprocessor-controlled prosthesis. This function that the proposed torque-sensitive actuator was key to enabling
could enable prosthesis users to walk on level ground indefinitely lightweight and efficient knee prostheses to replicate key biome-
without needing to charge the battery. chanical functions of the biological knee joint during ambulation.
On the basis of modeling, in step-over-step and double stair An articulated toe joint is necessary for a powered prosthesis to
ascent, the knee motor injected 0.49 ± 0.01 and 1.28 ± 0.09 J/kg replicate key biomechanical functions of the missing biological foot

Tran et al., Sci. Robot. 7, eabo3996 (2022) 23 November 2022 9 of 17

SCIENCE ROBOTICS | RESEARCH ARTICLE

(46, 47). However, adding a dedicated actuator for the toe joint in- biomechanical functions of the biological leg during walking, stair
evitably increases the overall prosthesis weight, which has well- ascent, and stair descent.
known negative effects on gait (72). Our experiments showed that In standard mode, the Utah Bionic Leg was expected to allow for
powering the ankle and the toe joint with a single actuator could 15,460 steps on a single battery charge. This number was higher
provide close to normative toe and ankle biomechanics in than the average steps taken by individuals with lower-limb ampu-
walking. As expected, using a single actuator results in a lighter tation [1500 (1)] and nonamputee individuals [7500 to 10,000 (2)].
and smaller design compared with using two actuators (51). Exper- Thus, the experiments suggest that the Utah Bionic Leg could
iments showed that an underactuated design could be highly effi- support multiple days of use on a single battery charge, like micro-
cient, because substantial mechanical energy was transferred from processor-controlled prostheses. However, note that kinematics, ki-
the toe joint to the ankle joint during ambulation (~3.3 J per stride; netics, and net energy injection change with walking speed in
table S7). In accordance with the simulations (Fig. 2), experiments nonamputee individuals (73). This speed-dependent behavior
with one amputee participant showed that locking the toe joint re- could be replicated in a powered prosthesis (74) and may have a sub-
sulted in a substantial increase in the required motor velocity (fig. stantial effect on electrical energy consumption and battery life.
S10) and mechanical power and consequently higher electrical When the Utah Bionic Leg was set to replicate the biomechanical
energy consumption. Moreover, the mechanical energy at the functions of a passive prosthesis with a microprocessor-controlled
ankle joint decreased likely due to the motor reaching the knee (Fig. 6) (75), the mechanical energy dissipation at the knee
winding limit. Thus, locking the joint had a substantial negative joint resulted in the Utah Bionic Leg regenerating 2.0 J of electrical
effect on the performance of the proposed ankle/foot prosthesis energy per stride. In this passive mode, a user could walk indefinite-
by spending more electrical energy to generate less mechanical ly even if the battery was depleted. This functionality is essential to
energy at the ankle. Like other powered prostheses, the proposed real-world viability, because users may not have access to a charger

Downloaded from https://www.science.org on March 31, 2024

underactuated mechanism uses a spring in series with the motor or may forget to charge the prosthesis. Thus, the Utah Bionic Leg
(Fig. 1). The proposed spring assembly is uniquely integrated offers functionality that is not currently available to individuals with
with the ball screw nut to achieve a lightweight (~125 g) and above-knee amputation.
compact design (30-mm outer diameter) that fits into the narrow Simulations show that increasing the range of motion of the
foot shell used for our prototype (40-mm width). The proposed torque-sensitive actuator would have resulted in better perfor-
compliant, underactuated mechanism is key to enable a lightweight mance. However, in the current knee implementation, the range
and efficient powered ankle/foot prosthesis to replicate key biome- of motion of the torque-sensitive joint is limited to keep the
chanical functions of the biological ankle and toe joint during overall knee dimensions comparable to a microprocessor-con-
ambulation. trolled knee, resulting in suboptimal performance (Fig. 1). To
Preclinical tests with three participants with an above-knee am- reduce the negative effects of the suboptimal range of motion, we
putation showed that the Utah Bionic Leg provided gait kinematics selected a relatively compliant spring, which causes the torque-sen-
and kinetics similar to nonamputee individuals (Fig. 6 and table S6). sitive joint to saturate before the maximum knee torque is achieved.
The toe joint worked as expected from simulation during walking, Because of the saturation, the spring deflection cannot be used to
and it achieved an even larger range of motion in stair descent, measure the knee torque. Therefore, we could not implement
which was not simulated. There were some visible differences closed-loop torque control. Benchtop tests against an external
between participants in early stance phase of walking. Participant load cell show that the performance of the open-loop knee torque
2 showed physiological stance knee flexion with kinematics and ki- controller is adequate for a powered prosthesis. However, this result
netics profiles closely matching the nonamputee references (Fig. 6). likely depends on the low friction and inertia of the current knee
In contrast, participant 1 showed a relatively small and constant implementation and may not generalize. In addition, the perfor-
stance knee flexion angle, and participant 3 kept the knee fully ex- mance of the open-loop torque controller may degrade with time
tended against the mechanical end stop throughout early stance. as the system friction and damping change.
Abnormal stance knee flexion is often observed in prosthetics The analytical model also shows that a torque-sensitive actuator
(56–58) and is due to habitual compensatory movements used by can store and release energy during ambulation, theoretically reduc-
individuals with an amputation to walk with their prescribed pros- ing the required mechanical power at the motor. However, in the
thesis rather than the controls or mechanics of the Utah Bionic Leg. current implementation, we use a small, lightweight spring com-
Furthermore, there were visible differences in the biomechanics of pared with series-elastic actuators with a similar output torque
the Utah Bionic Leg and that of nonamputee individuals. In stair (20 versus 200 to 500 g) (11, 56). As a result, energy storage and
ascent, the powered ankle prosthesis did not actively push off, release are negligible (fig. S9). Another limitation is that the pro-
leading to visible differences in ankle kinetics (Fig. 6). These differ- posed torque-sensitive actuator increases the torque ratio only in
ences were due to the controller in use (68), which did not imple- response to extension knee torque (Fig. 7). This unidirectional
ment ankle push-off, rather than the powered prosthesis mechanics. torque sensitivity does not have a negative effect on the knee per-
In stair descent, the powered prosthetic ankle provided lower torque formance because the biological knee flexion torque is substantially
than the biological ankle (Fig. 6). The ankle resistive torque was set smaller than the knee extension torque during ambulation
during pilot tests based on the feedback received from the partici- (Fig. 2A). However, there is no obvious way to achieve bidirectional
pants. Thus, the observed difference was due to subjective prefer- torque sensitivity with the proposed design. Therefore, the benefits
ence rather than the prosthesis mechanics. Despite these of the proposed torque-sensitive actuator are more substantial in
limitations, the preliminary clinical validation showed the potential joints like the biological knee or the ankle, which show a marked
of the proposed powered prosthesis to replicate the key bias in torque. Thus, there are limitations related to the dimensions,
energy efficiency, torque sensing, and sensitivity that can be

Tran et al., Sci. Robot. 7, eabo3996 (2022) 23 November 2022 10 of 17

SCIENCE ROBOTICS | RESEARCH ARTICLE

Downloaded from https://www.science.org on March 31, 2024

Fig. 7. Knee design. (A) Kinematic diagram of the torque-sensitive actuator. (B) Schematic representation of the torque-sensitive actuator highlighting the forces acting
on the torque-sensitive joint. (C) Knee torque ratio as a function of the knee angle and the knee extension torque. (D) Knee model highlighting the main mechanical
components. (E) Knee model highlighting the main electrical components. (F) Range of motion of the knee joint. (G) Range of motion of the torque-sensitive joint.

Tran et al., Sci. Robot. 7, eabo3996 (2022) 23 November 2022 11 of 17

SCIENCE ROBOTICS | RESEARCH ARTICLE

achieved with the proposed torque-sensitive actuator. These limita- amputee gait mechanics and improve the mobility of individuals
tions should be considered in future studies aiming to use the pro- with above-knee amputations in real life. The Utah Bionic Leg
posed torque-sensitive actuator. can enable scientists to study both the effects of energy injection
The proposed underactuated toe-ankle mechanism enables the and the effect of active control on amputee gait mechanics
development of a lighter and more efficient ankle/foot prosthesis without the confounding effect of prosthesis weight. Moreover, le-
compared with using two separate actuators. However, it cannot veraging its lightweight design, future studies using the Utah Bionic
provide independent control of the ankle and toe joint. Thus, a Leg could include elderly and dysvascular participants, who lack the
key limitation of the underactuated design is that the ratio strength and balance required to use heavier powered devices.
between the ankle and toe torque is fixed and cannot be changed Moreover, the Utah Bionic Leg satisfies basic requirements for
based on the user’s needs or preference. Moreover, even with an use at home. Thus, it may enable researchers to conduct studies
underactuated design, adding a toe joint increases the prosthesis outside the laboratory space, expanding the landscape of powered
weight (about 100 g). Comparative tests with and without the toe prosthesis research.
joint or with different ankle/toe torque ratios are necessary to
assess the effect of the toe joint on clinical outcomes. These clinical
studies are necessary to justify the use of an articulated toe joint. MATERIALS AND METHODS
Moreover, simulations show that a more compliant spring would Knee design
have improved dynamic performance and electrical efficiency. The proposed torque-sensitive actuator comprises two closed kine-
However, it would have required a longer spring, which would matic chains connected in parallel. The first kinematic chain has five
have reduced the range of motion of the ankle and toe joints. joints in closed configuration (P1R2R3P2R1P1R2R3P2R1), creating a
These limitations are inherent to the proposed underactuated toe- five-bar mechanism with two degrees of freedom, where R1 is the

Downloaded from https://www.science.org on March 31, 2024

ankle mechanism and should be considered in future designs. output knee joint, P1is the input joint powered by a linear actuator,
Similar to most microprocessor-controlled and powered ankle/ and P2 is the torque-sensitive joint (Fig. 7A). The torque-sensitive
foot prostheses, the proposed ankle design does not have actuation joint (P2) connects to a prismatic joint (P3) in the second kinematic
in the frontal plane. Adding the frontal plane actuation is likely to chain (R4P3R3P2R1), which is passively actuated by a tension spring
increase the size and weight of the prosthesis. However, it may also (Fig. 7B). The position of the torque-sensitive joint (P2) affects the
improve clinical outcomes, especially when walking on inclines and relationship between the force at the input joint (P1) and torque at
rough terrains. Powered emulators (76, 77) and prostheses (78, 79) the output joint (R1), effectively regulating the torque ratio of the
with passive or active frontal plane actuation have been developed to actuator (the ratio between output torque and input force).
study the effectiveness of frontal plane actuation. Future work Because of the connection with the spring-loaded joint (P3), the po-
should consider implementing frontal plane actuation based on sition of the torque-sensitive joint (P2) depends on the force on the
the outcome of these studies. input joint (P1). As shown in Fig. 7B, when the linear actuator pow-
This design validation study shows that the Utah Bionic Leg has ering the input joint (P1) pulls to generate extension torque at the
the potential to replicate the key biomechanical functions of the output joint (R1), the tension spring (P3) extends, causing the
missing biological leg for participants spanning large ranges of torque-sensitive joint (P2) to move away from the output joint
height (160 to 191 cm) and body weight (59 to 91 kg without the (R1), increasing the moment arm of the pulling force generated
prosthesis). Our results suggest that the performance of the Utah by the linear actuator. Thus, the torque ratio increases passively as
Bionic Leg is appropriate for clinical studies assessing biomechanics a function of the extension torque on the output knee joint
and clinical outcomes in a statistically significant number of partic- (Fig. 7C). Because of the nonlinear kinematics of the mechanism,
ipants. Leveraging the modularity of our powered prosthesis, these the torque ratio also depends on the knee angle (Fig. 7C). Moreover,
clinical studies should assess the contribution of the three powered because of the presence of a second degree of freedom (the torque-
joints—the knee, the ankle, and the toe—to amputee mobility and sensitive joint), the actuator’s torque ratio is not the inverse of the
subjective preference. velocity ratio. Thus, the proposed actuator conceptually differs from
By replicating key biomechanical functions of the missing bio- continuous variable transmissions, as shown by the analytical
logical leg, robotic leg prostheses have the potential to improve am- model described in Supplementary Methods.
bulation for millions of people living with an above-knee The proposed torque-sensitive actuator is implemented in a
amputation (9, 80). However, excessive weight, size, and short compact, lightweight powered knee prosthesis (Fig. 7D). The
battery life have prevented existing robotic leg prostheses from torque-sensitive actuator is fully integrated in a 7075-T6 aluminum
achieving clinical success. Here, we show that a torque-sensitive ac- frame, which provides structural support and protection for the me-
tuator can enable a small, lightweight motor to efficiently provide chanics and electronics, and also works as a heat sink. The prosthe-
the wide ranges of torques and speeds required for a knee prosthesis. sis actuator comprises a DC motor, a helical gear pair, and a ball
We also show that a compliant, underactuated system can concur- screw and nut assembly supported by two miniaturized linear
rently power the toe and ankle joint and also enable substantial me- guides. Passive degrees of freedom between the actuation and the
chanical energy regeneration. Combined, these design solutions frame allow for self-alignment of the actuation system. The linear
enable a robotic leg prosthesis to replicate the key biomechanical actuator is connected to the torque-sensitive joint through two par-
functions of the biological knee, ankle, and toe in the sagittal allel bars, which are joined by a steel shaft. The steel shaft slides into
plane and match the weight, size, and battery life of microproces- two parallel slots in the top knee structure and connects to the
sor-controlled prostheses. spring-actuated joint, which is implemented using a tension coil
By enabling the development of prostheses that are both light- spring. Movie S2 shows an animation of the torque-sensitive actu-
weight and powered, this study provides a scientific tool to study ation. The top knee structure connects to the user’s socket with a

Tran et al., Sci. Robot. 7, eabo3996 (2022) 23 November 2022 12 of 17

SCIENCE ROBOTICS | RESEARCH ARTICLE

titanium male pyramid adapter like in commercially available pros- third gear transfers power between the second gear and the motor
theses. Mechanical end stops limit the range of motion of the knee shaft. The range of motion of the ankle and toe joint is 40° (±20°)
joint to 0° to 120° (Fig. 7, F and G). The top knee structure contains and 45°, respectively (Fig. 8, F and G). Absolute magnetic encoders
an embedded electronic system that handles data processing of the measure the angles of the ankle and the toe joint. A linear potenti-
absolute encoders in the knee and torque-sensitive joint as well as of ometer measures the length of the spring. A custom instrumented
the inertial measurement unit and also provides an additional pyramid provides ground reaction force sensing (55). A nine-axis
digital/analog interface for additional sensing and control modes inertial measurement unit is integrated into the foot shell to deter-
such as electromyography (Fig. 7E) (80). The motherboard, mine the foot orientation in space. The sensor outputs are processed
located in the main knee structure, uses a microcontroller by a dedicated embedded electronic system, whereas another em-
(PIC32) to run control algorithms and to communicate with the bedded electronic system runs all the control routines and commu-
power electronic board where the motor current driver and motor nicates with the power electronic board, which hosts the motor
chokes are located. The motherboard can host an optional embed- driver. An embedded six-cell 1200-mAh lithium-ion battery
ded computer (Raspberry Pi 3+ compute module) to save data and powers the ankle/foot. The knee and ankle modules communicate
communicate via Wi-Fi to an external laptop for data telemetry. An using a serial peripheral interface, which runs inside the pylon so
embedded six-cell 1200-mAh lithium-ion battery powers the knee. that there are no electrical wires exposed (fig. S7). The proposed
The powered knee is designed on the basis of the ISO 10328 stan- ankle/foot is designed on the basis of the ISO 10328 standard (see
dard. More details on the mechatronic implementation are provid- Supplementary Methods). Our powered ankle/foot prosthesis lever-
ed in the Supplementary Methods. ages the proposed compliant, underactuated mechanism to satisfy
similar weight and size requirements to microprocessor-controlled
Ankle design ankle/foot prostheses and actively powers both the ankle and toe

Downloaded from https://www.science.org on March 31, 2024

The kinematics of the proposed underactuated system comprises joints during ambulation.
five joints in closed configuration (R1R2P1R3R4P1R2R3P2R1), creat-
ing a five-bar mechanism with two degrees of freedom. As shown in Modeling and dynamic simulations
Fig. 8 (A and B), the prismatic joint P1 is actuated by a linear series- We used a simulation framework to support the design of the
elastic actuator, concurrently generating torque at the ankle (R1) torque-sensitive actuator and the compliant, underactuated
and the toe joint (R4). Because the mechanical advantage of the system. The framework was used to solve the system dynamics
linear actuator with respect to the ankle joint is greater than that and calculate the motor torque, speed, voltage, and current neces-
of the toe joint (R1 R2 . R3 R4 ), the ankle torque ratio is consider- sary to generate physiological torque and velocity at the knee or
ably bigger than the toe torque ratio (Fig. 8C). In other terms, the ankle joint during ambulation (Fig. 2A). In addition, the framework
ankle torque is greater than the toe torque for any given force gen- was used to calculate the position, velocity, force, and torque on
erated by the linear series-elastic actuator. Because of the nonlinear each mechanical element of the design. The framework used a
kinematics of the proposed mechanism, the torque ratios depend on brute-force optimization approach, which explores the whole
the joint angles, although this dependency is marginal for the pro- design space within defined ranges of the kinematic and actuator
posed actuator configuration (see Supplementary Methods). A me- parameters (tables S2 and S3). The solutions were tested against
chanical end stop limits the range of motion of the toe joint so that it several hard limits to assess feasibility. The battery voltage defined
cannot pass the neutral angle. In this configuration, the kinematic a hard limit for the instantaneous motor torque/speed combination.
chain loses one degree of freedom, and the proposed actuation Moreover, the ability of the motor to dissipate heat defined a hard
system is not underactuated, so the ankle joint can be controlled limit for the continuous current and torque that an electrical motor
without affecting the toe joint. Moreover, a viscoelastic element can safely provide without overheating. We also included hard
returns the toe joint to a neutral position when no force is limits on force/torque/speed of other joints, which require bearings,
applied at the input joint, such as at the end of swing phase. linear guides, or ball screws. We simulated several brushless motors
The proposed compliant, underactuated mechanism is imple- that could fit within the desired prosthesis dimension. Moreover,
mented in a lightweight, compact ankle/foot prosthesis (Fig. 8D). for each motor, we simulated several winding configurations in
A linear series-elastic actuator transmits power to both the toe combination with different battery voltages, below and above the
joint and the ankle joint through the pivot joints implemented in nominal voltage specified in the motor datasheet. Last, we selected
the toe and shank frame. The linear series-elastic actuator comprises the motor/winding/battery combination with the smallest total
an electrical motor, a ball screw with integrated nut-spring assem- mass based on the estimates provided by the simulation framework.
bly, and a custom gearbox with bevel and helical gears. The ball The framework is presented in detail in Supplementary Methods.
screw is located inside the carbon fiber foot shell taken from a com- The simulation framework was essential to engineer the torque-sen-
mercially available ankle/foot prosthesis (Ottobock Meridium), sitive actuator and the compliant, underactuated system, because
whereas an electrical motor is located inside the shank frame, the torque ratio depends nonlinearly on both the output joint posi-
which is machined out of 7075-T6 aluminum and provides struc- tion and torque. Therefore, it might not have been possible to find
tural support, protection, and heat sink capacity. The custom the proposed design solutions based solely on the intuition of the
gearbox enables the electrical motor to remain in a fixed position designer.
with respect to the shank frame while the ankle and toe joint
move, using a set of three bevel gears. The first gear is coaxial to Benchtop testing
the ball screw axis of rotation and transmits power to the ball The knee and ankle/foot module were tested separately on the
screw and the second gear. The second gear spins freely and is bench. Each module underwent multiple experiments using either
coaxial with the pivot joint R3 on the shank frame (Fig. 8B). The a fully constrained or a partially constrained setup. For the fully

Tran et al., Sci. Robot. 7, eabo3996 (2022) 23 November 2022 13 of 17

SCIENCE ROBOTICS | RESEARCH ARTICLE

Downloaded from https://www.science.org on March 31, 2024

Fig. 8. Ankle design. (A) Kinematic diagram of the underactuated mechanism. (B) Schematic representation of the compliant, underactuated mechanism. (C) Ankle
torque ratio and ankle/toe leverage as a function of the ankle angle and the toe angle. (D) Ankle model highlighting the main mechanical components. (E) Ankle model
highlighting the main electrical components. (F) Range of motion of the ankle joint. (G) Range of motion of the toe joint.

constrained setup, both the proximal and distal links of the module level was tested five times. We recorded joint torque and motor
were mechanically grounded to a testing jig, which served to prevent current over time. To quantify the passive behavior of the knee
movement and provide reaction forces and moments. For the par- and ankle/foot modules, we performed manual backdriving exper-
tially constrained setup, one link of the module was mechanically iments with a partially constrained setup. For each module, with the
grounded to a testing jig, and the remaining link was free to motor turned off, an experimenter manually moved the output joint
rotate. During experiments that require readings of output joint through its range of motion in two sessions. In the first session, we
torque, a six-axis load cell (Sunrise Instruments, M3713D) was con- slowly backdrove each module (~ 0.1 Hz) to quantify the minimum
nected in series with the proximal portion of the module. backdriving torque. In the second session, we backdrove each
To quantify the position response of the knee and ankle module with frequencies continuously varying from 0.2 to 1.5 Hz.
modules, we performed closed-loop position step response experi- The joint torque and position data of this experiment were then
ments with a partially constrained setup. We tested with three levels used as inputs to the System Identification Toolbox in MATLAB,
of reference step position (5°, 10°, and 15°) for both modules, and which estimated the joint’s reflected damping and inertia parame-
each level was tested five times. We recorded joint angle over time. ters using a two-pole one-zero model.
To quantify the torque response of the knee and ankle/foot To demonstrate the behavior of the knee module’s torque-sensi-
modules, we performed open-loop torque step response experi- tive transmission in high-torque dynamic tasks, we generated bell-
ments with a fully constrained setup. We tested with four levels of shaped joint torque profiles used in the stance phase of reciprocal
reference step torque (±20 and ±40 Nm) for both modules, and each stair climbing, with a partially constrained setup. The peak torque

Tran et al., Sci. Robot. 7, eabo3996 (2022) 23 November 2022 14 of 17

SCIENCE ROBOTICS | RESEARCH ARTICLE

was scaled to 90 Nm, and we performed testing with the torque-sen- before data collection. Participant 3 did not have any experience
sitive joint unlocked and locked at its minimum position. For each before the data collection session. Thus, he was given a few
condition, we repeated the test five times. We recorded the joint minutes to practice before we started recording data and videos. Ad-
torque, joint angle, and motor current. In addition, to demonstrate ditional participant information is available in table S5. We pro-
the behavior of the knee module’s torque-sensitive transmission in cessed the testing data offline as explained in detail in
high-speed dynamic tasks, we commanded two minimum jerk po- Supplementary Methods.
sition profiles that represented those during walking and stair ascent To accurately assess electrical energy consumption, we per-
ambulation, with a partially constrained setup. We performed formed an additional test with participant 1. For this test, we con-
testing with the torque-sensitive joint unlocked and locked at its figured the Utah Bionic Leg to use a single Li-ion battery (six-cell,
maximum position. For each condition, we repeated the test five 1600 mAh). After calibration, current and voltage sensors were
times. We recorded the desired and measured joint position added directly at the battery leads. We enclosed the battery and
over time. sensors in a dedicated three-dimensionally printed plastic cover
To demonstrate the effect of the underactuated ankle-toe mech- connected to the back of the knee module. We acquired the
anism on the linear actuator, we commanded the ankle joint to gen- sensor outputs in real time using the embedded knee electronics
erate a 0.5-Hz sinusoidal position profile with a partially (fig. S11). The participant walked on the treadmill at 1.25 m/s in
constrained setup. We performed testing under two conditions: both standard mode and passive mode while we recorded the
with the toe joint locked and with the toe joint manually synchro- battery voltage and current. We calculated the electrical battery
nized with the ankle motion to minimize linear actuator movement power offline by multiplying the measured battery voltage and
by the experimenter. For each condition, we tested three times. We current. In addition, we calculated the electrical energy offline by
recorded joint positions and the movement of the linear actuator. integrating the electrical power over time for each stride. Then,

Downloaded from https://www.science.org on March 31, 2024

To quantify the stiffness of the series spring of the ankle/foot we normalized the electrical energy consumption by the partici-
module, we constructed a custom jig to load the spring subassembly pant’s body weight and calculated the average during standard
in both tension and compression. One end of the spring subassem- and passive mode separately.
bly was mechanically grounded to the jig and connected in series to
the six-axis load cell, and the other end was manually loaded by the Control architecture
experimenter through a four-bar linkage leverage. For each loading For all amputee experiments, we used ambulation controllers that
direction, we performed three loading-unloading cycles from 0 to were previously validated with above-knee amputee participants
400 N. We recorded the force and linear displacement of the [walking (74, 81), stair ascent (68), and stair descent (45)].
spring and used linear regression to estimate the spring stiffness. During the experiments, we switched between the ambulation con-
To quantify the thermal response of the knee and ankle/foot trollers manually using a graphical user interface. Although the am-
modules, we supplied a constant current through the motor with bulation controllers were designed to provide the powered
a fully constrained setup. We measured the device temperature prosthesis with different functionalities, they shared a common hi-
with a thermocouple (Mouser 410-315) placed on the prosthesis erarchical control architecture (fig. S15). At the higher level, there
frame, close to the motor housing. The thermocouple readings was a finite-state machine with two states: stance and swing. The
were confirmed using a thermal camera (FLIR ONE). We supplied transition rules between these two states were specific to each am-
two levels of current to the motor: the nominal current as specified bulation controller and can be found in previous publications (45,
in the motor datasheet for 60 min and a higher level of current for 68, 74, 81). Stance and swing used different control algorithms, im-
10 min. The actual current was verified using direct measurements plemented in the midlevel controller. In swing, we used a position
from the motor driver. We recorded prosthesis temperature and controller. The position controller defined a desired position either
used thermal modeling (see Supplementary Methods) to determine using a minimum jerk optimization (74) or based on the move-
the improved continuous and short-term current limits of the ments of the user’s residual limb (68). This approach allowed for
motor frame assembly for each module. smooth transitions between stance and swing in all conditions as
well as for indirect volitional control of the prosthesis behavior
Amputee testing protocol (81). In stance, we used a torque controller. The torque controller
We recruited three individuals with above-knee amputation (34 ± 6 defined a desired joint torque based on quasi-stiffness profiles (82)
years old, 72 ± 16 kg body mass; Fig. 5) for this study, and all par- extracted from able-bodied datasets of ambulation biomechanics
ticipants provided written consent to the University of Utah’s Insti- and implemented using lookup tables for maximal computational
tutional Review Board–approved clinical testing protocols. For data efficiency (74). This approach enabled the prosthesis to provide bio-
collection, the participants were fitted to the powered leg by a cer- logically accurate torque as the user walked at variable speeds and
tified prosthetist and then ambulated on level ground and stairs, all cadences (74) or climbed stairs with different stair heights or gait
in one test session. Level-ground walking was performed on the patterns (68) and required minimal (68) to no participant-specific
treadmill at a speed of 1.25 m/s, and stair ascent (both step-over- tuning (74). The desired torque or position was fed to dedicated
step and two steps at a time) and descent were performed with a low-level controllers, which generate current commands for the
reciprocal pattern. The participants had different levels of experi- knee and ankle motor drivers. For both the knee and ankle/foot
ence with the Utah Bionic Leg before data collection. Participant modules, the position controller was based on a simple proportional
1 had about 20 hours of experience with the Utah Bionic Leg as derivative (PD) compensator with additional feed-forward terms
he was involved in the early development of the device and for gravity and friction compensation (45). The knee module used
related ambulation controllers. Participant 2 had 2 hours of experi- an open-loop torque controller (fig. S14), which defined the desired
ence, gained during a familiarization session performed the week current based on an online estimate of the torque ratio using the

Tran et al., Sci. Robot. 7, eabo3996 (2022) 23 November 2022 15 of 17

SCIENCE ROBOTICS | RESEARCH ARTICLE

torque-sensitive model shown in Supplementary Methods as well as 15. T. Elery, S. Rezazadeh, C. Nesler, J. Doan, H. Zhu, R. D. Gregg, Design and benchtop vali-
dation of a powered knee-ankle prosthesis with high-torque, low-impedance actuators.
dynamic damping and inertia compensators similar to our previous
IEEE Int. Conf. Robot. Automat. 2008, 2788–2795 (2018).
powered prostheses (18, 45). In contrast, the ankle/foot module 16. A. H. R. Bellman, T. Sugar, R. D. Bellman, M. A. Holgate, T. G. Sugar, SPARKy 3: Design of an
used a closed-loop torque controller (fig. S14) with a PD compen- active robotic ankle prosthesis with two actuated degrees of freedom using regenerative
sator closing the loop on the online measurement of torque based kinetics. Proc. IEEE/RAS-EMBS Int. Conf. Biomedical Robotics and Biomechatronics 2008,
on the series spring deflection, as well as an online estimate of the 511–516 (2008).

ankle torque ratio based on the ankle and toe encoder readings. 17. S. K. Au, H. M. Herr, Powered ankle-foot prosthesis. IEEE Robot Autom Mag. 15, 52–59 (2008).
18. L. Gabert, S. Hood, M. Tran, M. Cempini, T. Lenzi, A compact, lightweight robotic ankle-foot
prosthesis: Featuring a powered polycentric design. IEEE Robot. Autom. Mag. 27,
Statistics 87–102 (2020).
For important quantitative metrics related to benchtop and 19. S. K. Au, J. Weber, H. Herr, Powered ankle–foot prosthesis improves walking metabolic
amputee participant testing, we calculated mean values (±SD). economy. IEEE Trans. Robot. 25, 51–66 (2009).
Within amputee testing data, each subsample represents one 20. Y. Feng, J. Mai, S. K. Agrawal, Q. Wang, Energy regeneration from electromagnetic induction
stride (segmented from heel strike to heel strike) from a given par- by human dynamics for lower extremity robotic prostheses. IEEE Trans. Robot. 36,
455–462 (2020).
ticipant and ambulation mode. To obtain fitted curves to character-
21. P. Cherelle, V. Grosu, A. Matthys, B. Vanderborght, D. Lefeber, Design and validation of the
ize the ankle/foot spring and thermal properties of the device, we ankle mimicking prosthetic (AMP-) foot 2.0. IEEE Trans. Neural Syst. Rehabilitation Eng. 22,
used the Curve Fitting Toolbox in MATLAB and obtained R 2 138–148 (2014).
values of at least 0.95. 22. P. Cherelle, V. Grosu, L. Flynn, K. Junius, M. Moltedo, B. Vanderborght, D. Lefeber, The ankle
mimicking prosthetic foot 3—Locking mechanisms, actuator design, control and experi-
ments with an amputee. Rob. Auton. Syst. 91, 327–336 (2017).
23. Ottobock Empower, 2019; https://shop.ottobock.us/Prosthetics/Lower-Limb-Prosthetics/
Supplementary Materials Feet---Microprocessor/Empower/p/1A1-2.

Downloaded from https://www.science.org on March 31, 2024

This PDF file includes:
24. Ossur, the POWER KNEE, www.ossur.com/en-us/prosthetics/knees/power-knee.
Supplementary Methods
Supplementary Discussion 25. J. Kim, J. Wensman, N. Colabianchi, D. H. Gates, The influence of powered prostheses on
Figs. S1 to S16 user perspectives, metabolics, and activity: A randomized crossover trial. J. Neuroeng.
Tables S1 to S7 Rehabil. 18, (2021).
26. E. S. Gardinier, B. M. Kelly, J. Wensman, D. H. Gates, A controlled clinical trial of a clinically-
tuned powered ankle prosthesis in people with transtibial amputation. Clin. Rehabil. 32,
Other Supplementary Material for this
319–329 (2018).
manuscript includes the following:
Movies S1 to S4 27. K. Lechler, B. Frossard, L. Whelan, D. Langlois, R. Müller, K. Kristjansson, Motorized bio-
mechatronic upper and lower limb prostheses—Clinically relevant outcomes. PM R 10,
S207–S219 (2018).
28. B. J. Hafner, R. L. Askew, Physical performance and self-report outcomes associated with
REFERENCES AND NOTES use of passive, adaptive, and active prosthetic knees in persons with unilateral, transfe-
1. R. Versluys, P. Beyl, M. van Damme, A. Desomer, R. van Ham, D. Lefeber, Prosthetic feet: moral amputation: Randomized crossover trial. J. Rehabil. Res. Dev. 52, 677–700 (2015).
State-of-the-art review and the importance of mimicking human anklefoot biomechanics. 29. J. D. Smith, P. E. Martin, Effects of prosthetic mass distribution on metabolic costs and
Disabil. Rehabil. Assist. Technol. 4, 65–75 (2009). walking symmetry. J. Appl. Biomech. 29, 317–328 (2013).
2. A. Hahn, I. Sreckovic, S. Reiter, M. Mileusnic, First results concerning the safety, walking, and 30. J. D. Smith, P. E. Martin, Short and longer term changes in amputee walking patterns due to
satisfaction with an innovative, microprocessor-controlled four-axes prosthetic foot. Pros- increased prosthesis inertia. J. Prosthet. Orthot. 23, 114–123 (2011).
thet. Orthot. Int. 42, 350–356 (2018). 31. Y. S. Narang, V. N. M. Arelekatti, A. G. Winter, The effects of prosthesis inertial properties on
3. J. W. Michael, Modern prosthetic knee mechanisms. Clin. Orthop. Rel. Res. 361, prosthetic knee moment and hip energetics required to achieve able-bodied kinematics.
39–47 (1999). IEEE Trans. Neural Syst. Rehabilitation Eng. 24, 754–763 (2016).
4. J. Thiele, C. Schöllig, M. Bellmann, M. Kraft, Designs and performance of three new mi- 32. A. Baimyshev, B. Lawson, M. Goldfarb, Design and preliminary assessment of lightweight
croprocessor-controlled knee joints. Biomed. Tech. (Berl) 64, 119–126 (2018). swing-assist knee prosthesis, in Proceedings of the 2018 40th Annual International Confer-
5. M. Goldfarb, Consideration of powered prosthetic components as they relate to micro- ence of the IEEE Engineering in Medicine and Biology Society (EMBC) (IEEE, 2018),
processor knee systems. J. Prosthet. Orthot. 25, P65–P75 (2013). pp. 3198–3201.
6. D. A. Winter, Energy generation and absorption at the ankle and knee during fast, natural, 33. T. Lenzi, J. Sensinger, J. Lipsey, L. Hargrove, T. Kuiken, Design and preliminary testing of the
and slow cadences. Clin. Orthop. Relat. Res. 175, 147–154 (1983). RIC hybrid knee prosthesis, in Proceedings of the Annual International Conference of the IEEE
7. B. J. McFadyen, D. A. Winter, An integrated biomechanical analysis of normal stair ascent Engineering in Medicine and Biology Society, EMBS (IEEE, 2015), pp. 1683–1686.
and descent. J. Biomech. 21, 733–744 (1988). 34. T. Lenzi, M. Cempini, L. Hargrove, T. Kuiken, Hybrid actuation systems for lightweight
8. T. Schmalz, S. Blumentritt, R. Jarasch, Energy expenditure and biomechanical characteris- transfemoral prostheses, in Proceedings of the Frontiers in Biomedical Devices, BIOMED -
tics of lower limb amputee gait. Gait Posture 16, 255–263 (2002). 2017 Design of Medical Devices Conference, DMD 2017 (ASME, 2017), pp. 1–2.
9. S. R. Wurdeman, P. M. Stevens, J. H. Campbell, Mobility analysis of AmpuTees (MAAT I): 35. T. Lenzi, M. Cempini, L. Hargrove, T. Kuiken, Design, development, and testing of a light-
Quality of life and satisfaction are strongly related to mobility for patients with a lower limb weight hybrid robotic knee prosthesis. Int. J. Robot. Res. 37, 953–976 (2018).
prosthesis. Prosthet. Orthot. Int. 42, 498–503 (2018). 36. M. K. Shepherd, E. J. Rouse, The VSPA foot: A quasi-passive ankle-foot prosthesis with
10. M. Goldfarb, B. E. Lawson, A. H. Shultz, Realizing the promise of robotic leg prostheses. Sci. continuously variable stiffness. IEEE Trans. Neural Syst. Rehabilitation Eng. 25,
Transl. Med. 5, 210ps15 (2013). 2375–2386 (2017).
11. E. J. Rouse, L. M. Mooney, H. M. Herr, Clutchable series-elastic actuator: Implications for 37. C. Lecomte, A. L. Ármannsdóttir, F. Starker, H. Tryggvason, K. Briem, S. Brynjolfsson, Variable
prosthetic knee design. Int. J. Rob. Res. 33, 1611–1625 (2014). stiffness foot design and validation. J. Biomech. 122, 110440 (2021).
12. S. Pfeifer, A. Pagel, R. Riener, H. Vallery, Actuator with angle-dependent elasticity for bio- 38. T. Lenzi, M. Cempini, J. Newkirk, L. J. Hargrove, T. A. Kuiken, A lightweight robotic ankle
mimetic transfemoral prostheses. IEEE ASME Trans. Mechatron. 20, 1384–1394 (2014). prosthesis with non-backdrivable cam-based transmission, in Proceedings of the IEEE In-
13. L. Flynn, J. Geeroms, R. Jimenez-Fabian, B. Vanderborght, N. Vitiello, D. Lefeber, Ankle-knee ternational Conference on Rehabilitation Robotics (IEEE Computer Society, 2017),
prosthesis with active ankle and energy transfer: Development of the CYBERLEGs alpha- pp. 1142–1147.
prosthesis. Rob. Auton. Syst. 73, 4–15 (2015). 39. E. M. Glanzer, P. G. Adamczyk, Design and validation of a semi-active variable stiffness foot
14. E. Martinez-Villalpando, H. Herr, Agonist-antagonist active knee prosthesis: A preliminary prosthesis. IEEE Trans. Neural Syst. Rehabilitation Eng. 26, 2351–2359 (2018).
study in level-ground walking. J. Rehabil. Res. Dev. 46, 361–373 (2009).

Tran et al., Sci. Robot. 7, eabo3996 (2022) 23 November 2022 16 of 17

SCIENCE ROBOTICS | RESEARCH ARTICLE

40. H. L. Bartlett, S. T. King, M. Goldfarb, B. E. Lawson, A semi-powered ankle prosthesis and 66. D. A. Bruening, K. M. Cooney, F. L. Buczek, Analysis of a kinetic multi-segment foot model
unified controller for level and sloped walking. IEEE Trans. Neural Syst. Rehabilitation Eng. part II: Kinetics and clinical implications. Gait Posture 35, 535–540 (2012).
29, 320–329 (2021). 67. R. Riener, M. Rabuffetti, C. Frigo, Stair ascent and descent at different inclinations. Gait
41. S. Gao, J. Mai, J. Zhu, Q. Wang, Mechanism and controller design of a transfemoral pros- Posture 15, 32–44 (2002).
thesis with electrohydraulic knee and motor-driven ankle. IEEE/ASME Trans. Mechatron. 26, 68. S. Hood, L. Gabert, T. Lenzi, Powered knee and ankle prosthesis with adaptive control
2429–2439 (2021). enables climbing stairs with different stair heights, cadences, and gait patterns. IEEE Trans.
42. T. Lenzi, M. Cempini, L. J. Hargrove, T. A. Kuiken, Design, development, and validation of a Rob. 38, 1430–1441 (2022).
lightweight nonbackdrivable robotic ankle prosthesis. IEEE/ASME Trans. Mechatron. 24, 69. C. Tudor-Locke, C. L. Craig, W. J. Brown, S. A. Clemes, K. de Cocker, B. Giles-Corti, Y. Hatano,
471–482 (2019). S. Inoue, S. M. Matsudo, N. Mutrie, J. M. Oppert, D. A. Rowe, M. D. Schmidt, G. M. Schofield,
43. WIPO IP Portal, Variable transmission for assistive prosthesis device (2020); https:// J. C. Spence, P. J. Teixeira, M. A. Tully, S. N. Blair, How many steps/day are enough? for adults.
patentscope.wipo.int/search/en/detail.jsf?docId=WO2020047043. Int. J. Behav. Nutr. Phys. Act. 8, 79 (2011).
44. T. Lenzi, M. Cempini, L. J. Hargrove, T. A. Kuiken, Actively variable transmission for robotic 70. E. G. Halsne, M. G. Waddingham, B. J. Hafner, Long-term activity in and among persons with
knee prostheses, in Proceedings of the 2017 IEEE International Conference on Robotics and transfemoral amputation. J. Rehabil. Res. Dev. 50, 515–530 (2013).
Automation (ICRA) (IEEE, 2017), pp. 6665–6671. 71. G. A. Pratt, M. M. Williamson, Series elastic actuators, in Proceedings of the IEEE International
45. M. Tran, L. Gabert, M. Cempini, T. Lenzi, A lightweight, efficient fully powered knee pros- Conference on Intelligent Robots and Systems (IEEE, 1995), pp. 399–406.
thesis with actively variable transmission. IEEE Robot Autom Lett. 4, 1186–1193 (2019). 72. R. C. Browning, J. R. Modica, R. Kram, A. Goswami, The effects of adding mass to the legs on
46. F. Mager, J. Richards, M. Hennies, E. Dötzel, A. Chohan, A. Mbuli, F. Capanni, Determination the energetics and biomechanics of walking. Med. Sci. Sports Exerc. 39, 515–525 (2007).
of ankle and metatarsophalangeal stiffness during walking and jogging. J. Appl. Biomech. 73. A. H. Hansen, D. S. Childress, S. C. Miff, S. A. Gard, K. P. Mesplay, The human ankle during
34, 1–19 (2018). walking: Implications for design of biomimetic ankle prostheses. J. Biomech. 37,
47. P. F. DeFrino, J. W. Brodsky, F. E. Polio, S. J. Crenshaw, A. D. Beischer, First metatarsopha- 1467–1474 (2004).
langeal arthrodesis: A clinical, pedobarographic and gait analysis study. Foot Ankle Int. 23, 74. T. Lenzi, L. Hargrove, J. Sensinger, Speed-adaptation mechanism: Robotic prostheses can
496–502 (2002). actively regulate joint torque. IEEE Robot. Autom. Mag. 21, 94–107 (2014).
48. E. C. Honert, G. Bastas, K. E. Zelik, Effect of toe joint stiffness and toe shape on walking 75. S. Hood, M. K. Ishmael, A. Gunnell, K. B. Foreman, T. Lenzi, A kinematic and kinetic dataset of
biomechanics. Bioinspir. Biomim. 13, 066007 (2018). 18 above-knee amputees walking at various speeds. Sci. Data 7, 150 (2020).

Downloaded from https://www.science.org on March 31, 2024

49. A. M. Grabowski, J. Rifkin, R. Kram, K3 promoter™ prosthetic foot reduces the metabolic 76. M. Kim, H. Lyness, T. Chen, S. H. Collins, The effects of prosthesis inversion/eversion stiffness
cost of walking for unilateral transtibial amputees. J. Prosthet. Orthot. 22, 113–120 (2010). on balance-related variability during level walking: A pilot study. J. Biomech. Eng. 142,
50. J. Zhu, Q. Wang, L. Wang, On the design of a powered transtibial prosthesis with stiffness 091011 (2020).
adaptable ankle and toe joints. IEEE Trans. Ind. Electron. 61, 4797–4807 (2014). 77. M. Kim, S. H. Collins, Step-to-step ankle inversion/eversion torque modulation can reduce
51. J. Zhu, H. She, Q. Huang, Pantoe II: Improved version of a powered transtibial prosthesis effort associated with balance. Front. Neurorobot. 11, 62 (2017).
with ankle and toe joints, in Proceedings of the Frontiers in Biomedical Devices, BIOMED - 78. E. A. Rogers, M. E. Carney, S. H. Yeon, T. R. Clites, D. Solav, H. M. Herr, An ankle-foot pros-
2018 Design of Medical Devices Conference (DMD, 2018), pp. 1–3. thesis for rock climbing augmentation. IEEE Trans. Neural Syst. Rehabilitation Eng. 29,
52. M. Laffranchi, N. Boccardo, S. Traverso, L. Lombardi, M. Canepa, A. Lince, M. Semprini, 41–51 (2021).
J. A. Saglia, A. Naceri, R. Sacchetti, E. Gruppioni, L. de Michieli, The Hannes hand prosthesis 79. E. M. Ficanha, G. A. Ribeiro, H. Dallali, M. Rastgaar, Design and preliminary evaluation of a
replicates the key biological properties of the human hand. Sci. Robot. 5, eabb0467 (2020). two DOFs cable-driven ankle-foot prosthesis with active dorsiflexion-plantarflexion and
53. Ottobock Genium X3, https://shop.ottobock.us/Prosthetics/Lower-Limb-Prosthetics/ inversion-eversion. Front. Bioeng. Biotechnol. 4, 36 (2016).
Knees---Microprocessor/X3/Genium-X3/p/3B5-3~5ST. 80. F. Zhang, H. Huang, Source selection for real-time user intent recognition toward volitional
54. Otto Bock Meridium Product Page, https://shop.ottobock.us/Prosthetics/Lower-Limb- control of artificial legs. IEEE J. Biomed. Health Inform. 17, 907–914 (2013).
Prosthetics/Feet---Microprocessor/Meridium/p/1B1. 81. J. Mendez, S. Hood, A. Gunnel, T. Lenzi, Powered knee and ankle prosthesis with indirect
55. L. Gabert, T. Lenzi, Instrumented pyramid adapter for amputee gait analysis and powered volitional swing control enables level-ground walking and crossing over obstacles. Sci.
prosthesis control. IEEE Sens. J. 19, 1–11 (2019). Robot. 5, eaba6635 (2020).
56. A. F. Azocar, L. M. Mooney, J. F. Duval, A. M. Simon, L. J. Hargrove, E. J. Rouse, Design and 82. E. J. Rouse, R. D. Gregg, L. J. Hargrove, J. W. Sensinger, The difference between stiffness and
clinical implementation of an open-source bionic leg. Nat. Biomed. Eng. 4, 941–953 (2020). quasi-stiffness in the context of biomechanical modeling. IEEE Trans. Biomed. Eng. 60,
57. B. E. Lawson, J. Mitchell, D. Truex, A. Shultz, E. Ledoux, M. Goldfarb, A robotic leg prosthesis: 562–568 (2013).
Design, control, and implementation. IEEE Robot. Autom. Mag. 21, 70–81 (2014).
58. T. Elery, S. Rezazadeh, C. Nesler, R. D. Gregg, Design and validation of a powered knee-ankle Acknowledgments: We acknowledge K. Rasmussen, MSPO, CPO for helping with prosthesis
prosthesis with high-torque, low-impedance actuators. IEEE Trans. Robot. 36, fitting and alignment, C. Buchanan for helping with the spring design and benchtop
1649–1668 (2020). characterization, and Ottobock for donating the carbon fiber foot keel used for the powered
59. I. Hunter, J. M. Hollerbach, J. Ballantyne, A comparative analysis of actuator technologies ankle/foot prosthesis. Funding: This work was partly funded by the National Institutes of Health
for robotics. Robot. Rev. , 299–342 (1992). through award no. R01HD098154, by the Department of the Defense through award no.
60. K. W. O’Brien, P. A. Xu, D. J. Levine, C. A. Aubin, H. J. Yang, M. F. Xiao, L. W. Wiesner, W81XWH2110037, and by the Rocky Mountain Center for Occupational and Environmental
R. F. Shepherd, Elastomeric passive transmission for autonomous force-velocity adaptation Health through NIOSH ERC grant no. T420H008414. Author contributions: T.L., M.T., and L.G.
applied to 3D-printed prosthetics. Sci. Robot. 3, eaau5543 (2018). conceived and implemented the prosthesis design. M.T., L.G., and S.H. performed the
experiments. M.T. and L.G. processed the experimental data and performed the analysis. All
61. F. Crenna, G. B. Rossi, M. Berardengo, Filtering biomechanical signals in movement analysis.
authors discussed the results and commented on the manuscript. T.L. obtained funding and
Sensors 21, 4580 (2021).
directed the research activities. Competing interests: The authors are coinventors on patents
62. J. Zhu, C. Jiao, I. Dominguez, S. Yu, H. Su, Design and backdrivability modeling of a portable
and disclosures pertaining to the results presented in the paper, one of which is licensed to
high torque robotic knee prosthesis with intrinsic compliance for agile activities. IEEE/ASME
Ottobock. T.L. is affiliated with the Rocky Mountain Center for Occupational and Environmental
Trans. Mechatron. 27, 1837–1845 (2022).
Health, for which he serves as the Acting Director of the Ergonomics and Safety Program. Data
63. DQYDJ, Weight Percentile Calculator for Men and Women in the United States; https:// and materials availability: All data needed to support the conclusions of this manuscript are
dqydj.com/weight-percentile-calculator-men-women. included in the main text or Supplementary Materials.
64. D. A. Winter, Biomechanics and Motor Control of Human Movement: Fourth Edition (John
Wiley & Sons, 2009). Submitted 5 February 2022
65. D. A. Bruening, K. M. Cooney, F. L. Buczek, Analysis of a kinetic multi-segment foot model. Accepted 31 October 2022
Part I: Model repeatability and kinematic validity. Gait Posture 35, 529–534 (2012). Published 23 November 2022
10.1126/scirobotics.abo3996

Tran et al., Sci. Robot. 7, eabo3996 (2022) 23 November 2022 17 of 17