Design and Experimental Characterisation of a Novel Quasi-Direct Drive Actuator for Highly Dynamic Robotic Applications

Abstract

This paper presents the design and experimental results of
a proprioceptive, high-bandwidth quasi direct drive (QDD)
actuator for highly dynamic robotic applications. A
comprehensive review of the mechanical design of the
actuator is presented, with particular focus on the design
parameters affecting the dynamic performance of the
actuator. After the description of the new PULSE115-60
actuator, a full specification is provided. Fundamental
parameters to describe the dynamic behaviour of an actuator
are discussed, and an experimental method to determine
speed and torque bandwidth of the actuator is presented. A
rigorous method to determine backdrive torque is also
explained. Finally, experimental results quantifying the
dynamic performance of the PULSE115-60 actuator are discussed.
The PULSE115-60 actuator has a highly dynamic response,
surpassing the torque bandwidth at low torque amplitudes
showcased in state-of-the-art literature. The differences
between current and torque bandwidth, two concepts often
conflated in literature, are elucidated. Experimental
procedures detailed in previous work are discussed and a
novel standardised procedure is proposed for robust
characterisation and fair comparison of different actuation
systems. Finally, performance results for PULSE115-60 are
presented, reaching a real torque bandwidth of 66.3 Hz at
an amplitude of 6 N·m, +-0.11º of backlash and 0.37 N·m of
backdrive torque.

Full text

Design and Experimental Characterisation of a Novel Quasi-Direct
Drive Actuator for Highly Dynamic Robotic Applications
C. Adri´
an P´
erez-D´
ıaz∗, Ignacio Mu˜
noz, Daniel Martin-Hern´
andez, Carlos Candelo-Zuluaga, Ivan Torres,
Jordi Mars`
a, Daniel Sanz-Merodio, and Miguel L´
opez
Abstract— This paper presents the design and experimen-
tal results of a proprioceptive, high-bandwidth quasi-direct
drive (QDD) actuator for highly dynamic robotic applications.
A comprehensive review of the mechanical design of the
PULSE115-60 actuator is presented, with particular focus on
the design parameters affecting the dynamic performance of
the actuator and a full specification is provided. Fundamental
parameters to describe the dynamic behaviour of an actuator
are discussed, and an experimental method to determine speed
and torque bandwidth of the actuator is presented. A rigorous
method to determine backdrive torque is also explained. Finally,
experimental results quantifying the dynamic performance of
the PULSE115-60 actuator are discussed. The PULSE115-60
actuator has a highly dynamic response, surpassing the torque
bandwidth at low torque amplitudes showcased in state-of-
the-art literature. The differences between current and torque
bandwidth, two concepts often conflated in literature, are
elucidated. Experimental procedures detailed in previous work
are discussed and a novel standardised procedure is proposed
for robust characterisation and fair comparison of different
actuation systems. Finally, performance results for PULSE115-
60 are presented, demonstrating a torque bandwidth of 66.3 Hz
at an amplitude of 6 N·m, ±0.11◦of backlash and 0.37 N·m of
backdrive torque.
Index Terms— Actuation and Joint Mechanisms, Dynamics,
Force Control, Bandwidth, Backdrivability, Transparency, Pro-
prioception, Quasi-Direct Drive, Development and Prototyping,
Physical Human-Robot Interaction
I. INT RODUCT IO N
Robot designers often face the difficult problem of de-
ciding whether to enhance the payload capacity of a robotic
system or maximise its dynamic performance. Two main con-
tributors are to blame for this design conundrum: actuators
(joints) and links.
Developing an actuator can be difficult since optimising
the design to maximise one performance parameter will
adversely affect others, forcing the designer to compromise.
For example, a direct or quasi-direct drive approach (using
actuators with low or no gear ratio) would provide higher
bandwidth and minimise the reflected inertia of rotary bodies
and friction at the expense of the maximum output torque
of the actuators and therefore mass specific torque (torque
to mass ratio) [1],[2],[3]. On the other hand, actuators with
a high gear ratio (harmonic, cycloidal, etc.) would certainly
provide a greater torque to weight ratio and a reduced trans-
mission backlash, at the expense of dynamic performance;
ARC Robotics, Arquimea Research Center, Edificio NANOTEC, Parque
Las Mantecas, Cmo de las Mantecas, S/N, 38320 La Laguna, Santa Cruz de
Tenerife, Spain. {cperez, imunoz, lmartin, ccandelo, ijtorres, jmarsa, dsanz,
mlopez}@arquimea.com
impact absorption [4], acceleration, bandwidth and friction
among others [5],[6].
The growing demand for robotic solutions for physical
Human-Robot Interaction (pHRI) has increased the need for
highly dynamic performance whilst ensuring an inherently
safe operation [7]. This can be achieved purely through
the mechanical design, using decoupling systems or load-
dependent and compliant mechanisms [8]; or through an
integral design philosophy where low inertia [9], high band-
width and proprioception are the main drivers for the actuator
design. This design paradigm requires a deep re-evaluation
of the design/selection choices for all the components of
an actuator, analysing the impact of each one on efficiency,
bandwidth and proprioceptive performance.
Fig. 1: Prototype of the PULSE115-60 actuator
In this paper, the PULSE115-60 actuator , shown in Fig.
1, is introduced as a first iteration following the QDD design
paradigm. Its design was performed from scratch to provide
outstanding dynamic response and proprioception capabili-
ties. In order to do so, a new electric motor (P100A) has
been developed with the objective of maximising torque per
watt and kilogram while operating at low speeds, and high
speed and torque bandwidth values. A new 5:1 transmission
has been also designed with the aim of reducing rotational
inertia and backlash for the specified operational envelope. In
the same way, a novel high stiffness torque sensor has been
developed and integrated within the actuator, to provide an
accurate measurement of both internal and external torques.
The low gear ratio also provides the possibility of estimat-
ing torque through current [10]. Two absolute, multi-turn
encoders have been integrated to provide angular position,
velocity and acceleration measurements for both the motor
and the output of the actuator.
pre-print ICRA24

PULSE115-60’s performance specifications, obtained
from analytical and experimental characterisation, are shown
in Table I.
Rated Torque 18.5 N·m @ 48VDC
Peak Torque 62.5 N·m @ 48VDC
Backdrive Torque 0.37 N·m
Transmission Type Planetary
Gear Ratio 5:1
Backlash ±0.11 deg
Max. Speed @ Rated Torque 90 rpm @ 48VDC
Max. Speed @ Peak Torque 36 rpm @ 48VDC
Mass 1.250 kg
Outer Diameter 113.8 mm
Width 64.2 mm
Hollow Shaft Diameter 10 mm
Rated Current 3.62 A
Peak Current 14.00 A
Voltage Bus 48 V
Rated Electric Power 170 W
Max. Electric Power 235 W
Mass Specific Torque 50 N·m/kg
Generalised Rated Torque Density 13.81 N·m/(A·kg)
Generalised Peak Torque Density 3.57 N·m/(A·kg)
Speed Bandwidth 87.4 Hz @100 rpm, 24VDC
Torque Bandwidth 66.3 Hz @ 6 N·m, 24VDC
Encoder 2x (20 bit) @ Input/Output
Hall Sensor 3x
Thermal Sensor 5x
Capacitive Torque Sensor 1x w/ 1DOF
TABLE I: PULSE115-60 Actuator Specification
II. ME CH AN ICAL DES IG N
The PULSE115-60 was designed to perform torque and
speed profiles within the limits displayed in Table I. The peak
torque and maximum speed values were defined according to
dynamic requirements extracted from bio-mechanic studies
of a human hip [11].
In order to achieve the desired performance, the design
aimed to maximise torsional stiffness and minimise inertia
of the mechanical transmission (also known as drivetrain),
inertia of the structural components and viscous damping.
[12]. The transmission gear ratio is another design parameter
which has a significant effect on the dynamic performance
due to its influence over reflected parameters at the output
[2], hence the selection of a low gear ratio. The following
equations show the relationship between the previously men-
tioned parameters:
KT=
N
∑
n=1
Kn
i2
n
(1)
BT=
N
∑
n=1
Bn
i2
n
(2)
JT=
N
∑
n=1
Jn
i2
n
(3)
Where KT,BT,JT, are torsional stiffness, viscous damp-
ing, and rotational inertia of the whole system evaluated
at the output respectively, and inis the transmission ratio
of each sub-system [12]. The connections and interactions
between each sub-system are considered in the calculations.
The torsional stiffness for metallic structure actuators is
usually not considered, since the effect of deformations of the
structural components are considered negligible compared to
the inertial and damping terms from a system response point
of view.
However, depending on the actuator architecture and/or the
overall transmission ratio, this term can become significant.
Therefore, actuator designers should take into account these
parameters and optimise them for a specific use case [2].
In order to facilitate optimisation, an actuator can be
understood as a set of sub-systems with multiple interfaces
between them: structural, electric motor, drivetrain, power
electronics, control electronics and sensing, as shown in Figs.
2 and 3. The design of each sub-system must take into
account not only its individual performance but also the in-
terfaces with other sub-systems to achieve the desired system
operation. For this reason mechanical and electro-mechanical
systems including a new capacitive torque sensor (CTS)
integrated in the PULSE115-60 actuator were designed from
scratch, to achieve the target performance.
Fig. 2: PULSE115-60 Exploded View
Fig. 3: PULSE115-60 Systems Diagram
A. Structural Design
The structural sub-system is comprised of four different
assemblies: a first structure, considered fixed; and three
structures which transmit the rotary motion of the motor to
the actuator output shaft.
pre-print ICRA24

The fixed structure is responsible for withstanding all
the loads generated by the actuator and reactions caused
by the environment. The structure is composed of a two
part housing. One of these parts (input housing) is rigidly
attached to an upstream body through the actuator mounting
interface. The second part (output housing) is fixed to the
first to enclose the remaining sub-systems of the actuator.
The output housing is the component onto which the CTS
and the motor stator are attached.
The torque sensor is rigidly attached to the output structure
and is capable of measuring sub-micron structural deforma-
tions to infer a torque value. All torque is transmitted through
the mechanical interfaces to the torque sensor.
As mentioned earlier, the actuator was designed to min-
imise mass while maximising the torsional stiffness at the
output, optimising the total bandwidth of the system. How-
ever, some torsional deformation is necessary for the torque
sensor to be able to measure accurately. Thus, a compromise
must be made between the overall torsional stiffness and the
proprioceptive capability provided by an integrated torque
sensor.
Fig. 4: Motor Stator with Integrated Geared Transmission
(a), Motor Rotor Attached to Sun Gear (b)
The first rotating sub-assembly, composed of the rotor,
motor magnets, rotor connector and sun gear, is supported
by two precision bearings. The rotor connector is rigidly at-
tached to the rotor of the motor and to the axially concentric
sun gear as shown in Fig. 4b.
The planet gear sub-assembly, the second rotating struc-
ture, is supported radially by a needle bearing and axially
by two thrust bearings that reduce friction and consequently
damping, increasing backdrivability.
The remaining rotating structure is composed of the carrier
and the counter carrier, supported with two ball bearings and
connected by the planet gear shafts. The whole assembly is
locked axially by means of three screws.
B. Electric Motor
The mechanical power of electromagnetic rotary actua-
tors is commonly provided by high-speed/low-torque elec-
tric motors. These actuators require the integration of a
mechanical transmission with a high reduction ratio to
adapt the mechanical power to the application requirements.
Designs based on this criteria display high precision and
high mass specific torque, however, system responsiveness
and proprioceptiveness are drastically reduced, among other
performance losses.
PULSE115-60 is equipped with a tailor-made outrunner
3-Phase PMSM electric motor (P100A), designed to provide
high toque at low speed. Table II displays its main specifica-
tions. Electric motors designed according to this paradigm
are known as torque motors. Integrating these motors in
actuators allows for low transmission ratios, lowering the
reflected inertia and damping, resulting again in increased
bandwidth.
DC Bus Voltage 48.00 V
Rated Current 3.65 A
Stall Current 14.00 A
Rated Torque 3.70 N·m
Stall Torque 12.50 N·m @ 48V DC BUS
Rated Torque Constant (KTr ) 1.00 N·m/A
Stall Torque Constant (KT s ) 0.90 N·m/A
Rated Speed 445 rpm
Max. Speed 500 rpm
Mass (Frameless) 590g
Volume Specific Torque 144.00 kN/m3@ Stall
Mass Specific Torque 14.70 N·m/kg @ Stall
Efficiency 79.00 % @ Rated
TABLE II: P100A Motor Specification
C. Drivetrain
Robotic platform operation can be divided into two
types: non-contact and contact. Non-contact operations (e.g.
performing tasks such as spray painting or welding) are
extensively utilised and can be widely found in industry.
The mechanical design of these kinds of robots requires
highly rigid links and high ratio transmissions, facilitating
precise and repeatable positioning [13]. In the same way,
one of the challenges in robotics is achieving high mass
specific torque [2], however this design approach harms
dynamic response since the platform loses impact mitigation
capability and system responsiveness, and therefore, torque
and speed bandwidth. These parameters are crucial for pHRI
and mobile robotics, thus, high mass specific torque actuators
may perform worse in dynamic applications.
The PULSE115-60 transmission was designed targeting
low inertia and low backlash while minimising torque ripple.
Moreover, backlash is a non-linearity that may considerably
affect the actuator’s control and mechanical performance
[14], hence decreasing its bandwidth. Besides, backlash
magnitude varies with temperature, lubrication, load and
transmission and gear meshing configuration, where each
individual gear meshing state combines to produce a different
backlash magnitude. The transmission was also designed
to have reduced peak-to-peak transmission error (PPTE)
[15] which reduces torque signal noise and enhances torque
bandwidth when the signal is used in the control loop.
As shown in Fig. 3 the actuator’s drive train is composed
of the electric motor fed by a DC power supply connected to
a an electric inverter that delivers the electrical signal defined
by the control output to the electric motor’s stator, which
converts the electrical power into mechanical power to the
pre-print ICRA24

rotor. This mechanical power is then directly transferred to
an epicyclic planetary transmission with a 5:1 gear ratio and,
consequently, to the actuator’s mechanical output interface.
The transmission is fitted inside the motor stator bore,
allowing for a compact actuator assembly. Geared planetary
transmissions have showcased superior performance com-
pared to other transmissions, such as strain wave drives,
in backdrivability, efficiency and associated motor current
consumption [16].
D. Capacitive Torque Sensor
Proprioceptive and high bandwidth actuators require inte-
grated force-feedback sensors among others. Capacitive tech-
nology has showcased better performance when compared
with strain gauges in robotics applications for measuring
mechanical displacements to infer torque [17].
The actuator is equipped with an integrated capacitive
torque sensor, with high torsional stiffness to enhance band-
width, where measured torque is inferred from capacitance
changes. Results presented in Fig. 5 show how the actuator
is capable of sensing a wide range of torques. Displayed data
have not been treated or post-processed and future work will
aim to improve the sensor’s performance.
Fig. 5: Measured Torques with Integrated Capacitive Torque
Sensor
III. CON TROL AND BA NDW IDTH EXP ER IM EN TS
The precise measurement of bandwidth and backdrive
torque and their use as indicators of dynamic performance in
actuators is a subject that has not been rigorously investigated
in modern robotics. This section presents the implementation
of an actuator control scheme, and details the experiments
conducted to determine the bandwidth and backdrive torque
of the presented actuator.
A. Backdrive Torque Experiment
The backdrive torque is defined as the minimum torque
required at the output to overcome the actuator’s internal
mechanical impedance and initiate the electric motor’s rotor
movement. It quantifies the ability of an actuator to be
driven from the load side. Low backdrive torque actuators
are essential for developing human-centred robotic devices.
To characterise the backdrivability of the PULSE115-60
actuator, the output shaft was moved with an external motor,
providing a sinusoidal speed signal with a frequency of 1 to
2 Hz, and the actuator motor phases left in open circuit. Pre-
vious studies detail methods for measuring backdrive torque.
One such method defines the backdrive torque as the torque
applied at which the output reaches a particular angular
velocity [18]. The angular velocity is selected somewhat
arbitrarily and therefore does not give a robust and reliable
measure of backdrive torque, making meaningful comparison
between actuators difficult.
In this paper, a rigorous approach is proposed for ex-
perimentally determining the backdrive torque of a rotary
actuator. The angle at the input and output of the actuator
are measured, and the ratio between their respective angular
velocities are calculated. Since the actuator is not backlash-
free, the backdrive torque is defined as the torque at which
the ratio between the velocities recorded by the input and
output encoders converges within ±1% error of the design
value (5:1). This criterion also takes into account transmis-
sion error inherent in geared drives.
B. Speed and Torque Control
This section presents the Field-Oriented Control (FOC)
scheme which manages the P100A electric motor integrated
in the PULSE115-60 actuator. Fig. 6 depicts the speed
and torque control schemes implemented and used in the
experiments to determine the actuator’s bandwidth response.
Fig. 6: Torque Control (a) and Speed Control (b) Schemes
Based on Field Oriented Control (FOC)
C. Bandwidth Calculation Experiments
Previous works present different criteria for defining the
torque bandwidth of an actuator. When measuring the band-
width of mechanical systems, the influence of the imple-
mented control should be taken into account.
For instance, when a current control loop is nested in
a torque loop [19], the torque bandwidth is one order
of magnitude lower when compared against a standalone
current loop inferring the transmitted torque from the current
measurement [2]. Additionally, when a standalone current
loop is used, if instead of being inferred the torque is mea-
sured by a transducer and fed back, a delay is included due
to the intrinsic bandwidth of the torque sensor, decreasing
torque bandwidth.
pre-print ICRA24

This paper presents a standalone current loop scheme, with
a torque transducer at the actuator output, used only for
bandwidth measurement as shown in 6 (a). This configu-
ration is chosen because the measurement of bandwidth will
include the lag time effects of the whole mechanical system,
which are not taken into account using other methods. This
measurement method allows for more realistic comparison
between actuator designs.
In addition, many of the experiments described in literature
estimate bandwidth from the response to a step stimulus
[18],[2]. Although it theoretically provides an adequate value
for the bandwidth, great precision in measurements is nec-
essary to obtain it accurately. Furthermore, calculating the
bandwidth of an actuator with a step-response experiment
is unreliable due to the inherent non-linear effect of back-
lash which significantly influences the achievable bandwidth
during reversing torque operation.
The stimulus for a robust characterisation of bandwidth
in this type of transmission must include changes in the
direction of movement, for example, a sinusoidal signal. In
reference [20], a method to measure torque bandwidth in
electric rotary actuators is presented. A set of experiments is
conducted to obtain the bandwidth of the actuator’s response
based on a set of constant amplitude chirp signals.
Fig. 7 presents a schematic layout of the test bench
designed for the experiments, Fig. 8 shows the finished
mechanical assembly. A SpeedGoat real-time target machine
provides the capacity to carry out the control presented in
the previous section while collecting the experimental data.
Fig. 7: Test Bench Utilised in the Characterisation Experi-
ments
Fig. 8: Torque Bandwidth Experimental Setup
In order to establish the bandwidth of the actuator, the ex-
perimental procedure followed is similar to the one outlined
in reference [20]. The actuator is excited with a sine wave
signal of fixed amplitude. Over time, the signal frequency is
increased in discrete steps. At each frequency, the response
of the actuator is measured. The actuator’s attenuation and
phase shift in response to the reference signal are calculated
at each frequency. The actuator bandwidth corresponds to
either the frequency at which the response is attenuated
below -3dB or the frequency at which the phase shift exceeds
180º.
The rotor shaft is mechanically fixed for the torque exper-
iments while left unrestricted for speed experiments.
A tuning process is used to find an appropriate combina-
tion of control constants Kpand Ki. The procedure starts by
determining the theoretical constants based on the physical
characteristics of the actuator, similar to the process in
reference [22], subsequently an algorithm performs a sweep
to test multiple combinations of Kpand Kito find the best
torque bandwidth.
D. Experimental Results and Data Analysis
Although speed and torque bandwidth are intrinsically
related, the benefits of either torque or speed bandwidth will
be exploited depending on the application. For example, a
wide speed bandwidth is necessary in drone flight control
systems and target tracking systems, while in industrial
robotic arms and rehabilitation robots, torque bandwidth
takes precedence.
1) Backdrivability Experiments:Backdrive torque of the
PULSE115-60 actuator is measured using the experimental
method specified in Section III-A, resulting in a value of
0.37 N·m.
Figure 9 shows the backdrive torque measurements for 4
out of the 15 different output angular positions evaluated,
spread evenly over one full output rotation. The backdrive
torque was calculated as the mean value of the data points
obtained. Consequently, different gear meshing configuration
effects were taken into account, including backlash and
transmission error variation during the whole rotation.
Fig. 9: PULSE115-60 Backdrive Torque Experiment Data
2) Speed Bandwidth Experiments:The experimental
measurements demonstrate that a speed bandwidth of 87.4
Hz at an amplitude of 100 rpm is achieved, comparable to
what is found in the literature, which reports a bandwidth of
5 Hz at an amplitude of 500 rpm [23]. Fig. 10 shows the
sweep results of the constants Kp and Ki used in the speed
PID loop and the bandwidth achieved.
pre-print ICRA24

Fig. 10: PULSE115-60 Speed Bandwidth vs Kp and Ki with
an Amplitude of 100 rpm
3) Torque and Current Bandwidth Experiments:To
demonstrate the difference between current control loop
bandwidth and torque bandwidth, both values were obtained
from the same experimental setup and run.
Fig. 11: PULSE115-60 Torque Bandwidth vs Torque Kp and
Ki with an Amplitude of 6 N·m
PULSE115-60 achieves a torque bandwidth of 66.3 Hz
at an amplitude of 6 N·m as shown in Fig.11, while, a
current bandwidth of 387 Hz is measured in the same
experiment as shown in Fig. 12. These measurements show
the importance of including the lag time effects of the whole
mechanical system to enable realistic comparison between
actuator designs.
Table III compares the dynamic characteristics of the most
commonly used actuators in robotics. It is worth mentioning
that the PULSE115-60 actuator reaches a bandwidth com-
parable to the QDD actuator results presented in reference
Fig. 12: PULSE115-60 Current Bandwidth vs Kp and Ki
with an Amplitude of 0.6 A
[1].
Nevertheless, it remains unclear if the bandwidth values
are truly comparable since the experimental procedure pre-
sented by Yu et al. [1] does not specify if the value measured
represents actual torque bandwidth or current loop bandwidth
as explained earlier.
IV. CON CLUSI ON S
In conclusion, this paper introduces PULSE115-60, the
first iteration of a new QDD actuator design with proprio-
ceptive capabilities, and enhanced dynamic bandwidth and
backdrivability.
A new set of experimental procedures are proposed to
characterise bandwidth and backdrivability in a repeatable
and accurate manner. Discussion clarifying the distinction
between torque and current bandwidth has been also pre-
sented, highlighting that toque bandwidth is a more robust
and pragmatic measure of actuator dynamic response.
Finally, the results obtained demonstrate how the
PULSE115-60 actuator design showcases excellent backdriv-
ability and bandwidth performance when compared popular
commercial actuators and state-of-the-art results available
in literature. However, effective comparison is hindered by
inconsistent and obscure measurement procedures.
ACKNOWLEDGEMENT
Funded by the European Union. Views and opinions
expressed are however those of the author(s) only and do
not necessarily reflect those of the European Union. Neither
the European Union nor the granting authority can be held
responsible for them.
Parameter EC45/HD[1] EC90/HD[1] Yu et al.[1] HEBI X8-16[21] PULSE115-60
Peak Torque [N·m] 8 40 17.5 34.0 62.5
Actuator Mass [kg] 0.5 1.8 0.77 0.49 1.25
Mass Specific Torque [N·m/kg] 16 22.2 20.7 77.6 50
Torque Bandwidth [Hz] 5.1 @ 5 N·m 4.2 @ 5 N·m 62.4 @ 10 N·m 19 @ 4 N·m 66.3 @ 6 N·m
Backdrive Torque [N·m] 2.88 6.1 0.97 - 0.37
Tansmission Ratio 50:1 100:1 8:1 1462.2 : 1 5:1
TABLE III: Actuator Dynamics Comparison
pre-print ICRA24

REF ER EN CE S
[1] S. Yu, T.-H. Huang, X. Yang, C. Jiao, J. Yang, Y. Chen, J. Yi, and
H. Su, “Quasi-direct drive actuation for a lightweight hip exoskeleton
with high backdrivability and high bandwidth,” IEEE/ASME Transac-
tions on Mechatronics, vol. 25, no. 4, pp. 1794–1802, 2020.
[2] P. M. Wensing, A. Wang, S. Seok, D. Otten, J. Lang, and S. Kim,
“Proprioceptive actuator design in the mit cheetah: Impact mitigation
and high-bandwidth physical interaction for dynamic legged robots,”
Ieee transactions on robotics, vol. 33, no. 3, pp. 509–522, 2017.
[3] A. Singh, N. Kashiri, and N. Tsagarakis, Design of a Quasi-Direct-
Drive Actuator for Dynamic Motions, Nov. 2020, vol. 64, journal
Abbreviation: Proceedings Publication Title: Proceedings.
[4] C. Rela˜
no, D. Sanz-Merodio, M. Lopez, and C. Monje, “Generaliza-
tion of impact response factors for proprioceptive collaborative robots,”
in Proceedings. 2023 IEEE International Conference on Robotics and
Automation. IEEE, 2023.
[5] E. Dunwoodie, R. Mutlu, B. Ugurlu, M. Yildirim, T. Uzunovic, and
E. Sariyildiz, “A High-Torque Density Compliant Actuator Design
for Physical Robot Environment Interaction,” in 2020 IEEE 16th
International Workshop on Advanced Motion Control (AMC), Sep.
2020, pp. 1–6, iSSN: 1943-6580.
[6] Q. Hua, W. Zhou, S. Zhu, Y. Yao, C. Cheng, A. Xie,
and D. Zhang, “Design of a high-torque robot joint and
its control system,” Journal of Physics: Conference Series,
vol. 2281, no. 1, p. 012007, jun 2022. [Online]. Available:
https://dx.doi.org/10.1088/1742-6596/2281/1/012007
[7] H. Lee, B. Lee, W. Kim, M. Gil, J. Han, and C. Han, “Human-
robot cooperative control based on pHRI (Physical Human-Robot
Interaction) of exoskeleton robot for a human upper extremity,”
International Journal of Precision Engineering and Manufacturing,
vol. 13, no. 6, pp. 985–992, Jun. 2012. [Online]. Available:
https://doi.org/10.1007/s12541-012-0128-x
[8] E. Saerens, R. G. Furn´
emont, J. Legrand, S. Crispel, P. Lopez Garcia,
T. Verstraten, B. Vanderborght, and D. Lefeber, “Novel specta
actuator to improve energy recuperation and efficiency,” Actuators,
vol. 11, no. 3, p. 64, Feb 2022. [Online]. Available:
http://dx.doi.org/10.3390/act11030064
[9] K. Urs, C. E. Adu, E. J. Rouse, and T. Y. Moore, “Alternative metrics
to select motors for quasi-direct drive actuators,” 2022.
[10] H. Xiang, K. Xiang, Y. Fang, M. Pang, B. Tang, J. Luo, and G. Zhu,
“Torque estimation base on quasi-direct drive actuators,” pp. 1637–
1642, 2022.
[11] K. Embry, “The Effect of Walking Incline and Speed
on Human Leg Kinematics, Kinetics, and EMG,” Oct.
2018. [Online]. Available: https://ieee-dataport.org/open-access/effect-
walking-incline-and-speed-human-leg-kinematics-kinetics-and-emg
[12] P. Lopez Garcia, E. Saerens, S. Crispel, A. Varadharajan, D. Lefeber,
and T. Verstraten, “Factors influencing actuator’s backdrivability
in human-centered robotics,” MATEC Web of Conferences, vol.
366, p. 01002, 2022. [Online]. Available: https://www.matec-
conferences.org/10.1051/matecconf/202236601002
[13] B. Eberman and J. Jr, “Whole-arm manipulation: Kinematics and
control,” p. 8, 03 1995.
[14] S. Habibi, J. Roach, and G. Luecke, “Inner-Loop Control for
Electromechanical (EMA) Flight Surface Actuation Systems,”
Journal of Dynamic Systems, Measurement, and Control,
vol. 130, no. 5, p. 051002, 08 2008. [Online]. Available:
https://doi.org/10.1115/1.2936382
[15] O. Mohammed, “A study of different considerations to meet gear
design requirements,” Procedia Structural Integrity, vol. 42, pp. 1607–
1618, 12 2022.
[16] M. J. Nandor, M. Heebner, R. Quinn, R. J. Triolo, and N. S. Makowski,
“Transmission comparison for cooperative robotic applications,”
vol. 10, no. 9. [Online]. Available: https://www.mdpi.com/2076-
0825/10/9/203
[17] M. Y. Cao, S. Laws, and F. R. y. Baena, “Six-axis force/torque sensors
for robotics applications: A review,” IEEE Sensors Journal, vol. 21,
no. 24, pp. 27 238–27 251, 2021.
[18] H. Zhu, C. Nesler, N. Divekar, M. Ahmad, and R. Gregg, “Design
and Validation of a Partial-Assist Knee Orthosis with Compact,
Backdrivable Actuation,” vol. 2019, Jun. 2019, pp. 917–924.
[19] T. Zhang and H. Huang, “A Lower-Back Robotic Exoskeleton:
Industrial Handling Augmentation Used to Provide Spinal
Support,” IEEE Robotics & Automation Magazine, vol. 25,
no. 2, pp. 95–106, Jun. 2018. [Online]. Available:
https://ieeexplore.ieee.org/document/8360955/
[20] J. Malzahn, N. Kashiri, W. Roozing, N. Tsagarakis, and D. Caldwell,
“What is the torque bandwidth of this actuator?” in 2017 IEEE/RSJ
International Conference on Intelligent Robots and Systems (IROS).
Vancouver, BC: IEEE, Sep. 2017, pp. 4762–4768. [Online]. Available:
http://ieeexplore.ieee.org/document/8206351/
[21] I. P. HEBI Robotics, “Hebi robotics: Documentation.” [Online].
Available: https://docs.hebi.us/hardware.htmlx-series-actuators
[22] M. G ˘
aiceanu, “Advanced Control of the Permanent Magnet Syn-
chronous Motor,” in Electric Machines for Smart Grids Applications -
Design, Simulation and Control, A. El-Shahat, Ed. IntechOpen, Dec.
2018.
[23] Y. Li, X. Guan, X. Han, Z. Tang, K. Meng, Z. Shi,
B. Penzlin, Y. Yang, J. Ren, Z. Yang, Z. Li, S. Leonhardt,
and L. Ji, “Design and Preliminary Validation of a Lower
Limb Exoskeleton With Compact and Modular Actuation,” IEEE
Access, vol. 8, pp. 66 338–66 352, 2020. [Online]. Available:
https://ieeexplore.ieee.org/document/9057535/
pre-print ICRA24