Cobotic Architecture for Prosthetics
Eric L. Faulring, J. Edward Colgate and Michael A. Peshkin
Abstract— We envision cobotic infinitely-variable transmissions (IVTs) as an enabling technology for haptics and prosthetics that will allow for increases in the dynamic range of
these devices while simultaneously permitting reductions in
actuator size and power requirements. Use of cobotic IVTs
eliminates the need to make compromises on output flow and
effort, which are inherent to choosing a fixed transmission ratio
drivetrain. The result is a mechanism with enhanced dynamic
range that extends continuously from a completely clutched
state to a highly backdrivable state. This high dynamic range
allows cobotic devices to control impedance with a high level of
fidelity. In this paper, we discuss these and other motivations
for using parallel cobotic transmission architecture in prosthetic
devices.
I. INTRODUCTION
Four key requirements of robotic technologies used for
prosthetics, orthotics and rehabilitation robotics are low
weight, low energy consumption, safety and controllability.
We propose cobotic technology as a transmission architecture
that can address all of these issues. Given a set of design
criteria for a multi-degree-of-freedom mechanism, such as
maximum flow, maximum effort and maximum power, we
find that a cobot can meet these requirements with reduced
numbers of high power actuators, reduced size requirements
for those actuators and increased power efficiency relative to
conventional actuation systems [1].
Cobots are robots that utilize the nonholonomic constraints
of steered wheels in order to relate the relative velocities
of mechanism links. These steered wheels form the basis
of a cobotic infinitely variable transmission (IVT). Such a
transmission can be smoothly adjusted between an infinite
reduction and a zero reduction. Cobotic IVTs have been
developed that relate two translational velocities [5], [6],
two rotational velocities [3], or a rotational velocity to a
translational velocity [1], and have been utilized in many
prototype devices. The use and control of nonholonomic
constraints (rolling elements) as the basis for passive cobot
technology is best summarized by Peshkin et al. [4] and
Gillespie et al. [2].
Cobotic technology provides a highly power efficient
and weight efficient transmission architecture that can have
minimal dissipation and trivial dynamics. Gear trains, timing
belt transmissions, hydraulic and pneumatic systems as well
as cable systems all have dissipative losses that result in
This work was supported by the DOE grant number DE-FG0701ER63288.
E.L. Faulring is with Chicago PT, LLC, 2510 Gross Point Road, Evanston,
IL, 60201, USA eric.faulring@ieee.org
J.E. Colgate and M.A. Peshkin are with the Mechanical Engineering
Department, Northwestern University, 2145 Sheridan Road, Evanston, IL,
60208, USA colgate,peshkin@northwestern.edu
heat and noise generation. In addition, stiction, friction,
compliance and backlash in these transmissions add highly
nonlinear dynamics to mechanisms. Cobotic transmissions
utilizing bearing quality steel components in dry-friction
rolling-contact have none of these nonlinearities. Haptic simulations have unusual realism when displayed on the Cobotic
Hand Controller, a six-degree-of-freedom haptic display [1].
The crisp distinction between free and forbidden directions
of motion is a salient feature of cobots. This performance
does not arise from elaborate control algorithms, but from
the inherent physical characteristics of the device due to
the utilization of nonholonomic rolling constraints in its
transmissions.
In the remainder of this paper we discuss the potential
weight, energy, dynamic range and safety benefits of cobotic
architecture for prosthetics.
II. WEIGHT SAVINGS
Cobotic architecture can lead to significant weight savings,
since an infinitely variable transmission paired with a small
motor can achieve speeds and loads of a much larger motor
with only a fixed reduction. While such a design is still
constrained by the maximum mechanical power rating of
the motor, this is an acceptable compromise in prosthetics
where the extreme operating points of maximum force with
minimal velocity and maximum velocity with minimal force,
both low mechanical output power conditions, are almost
exclusively employed. Using an infinitely variable cobotic
transmission can eliminate the need to make compromises
on output flow and effort, which are inherent in choosing a
fixed transmission ratio. Since an infinite reduction can be
obtained, the active lift limitation of the actuator is secondary
to the speed rating and power rating.
Additional weight savings can be realized by parallel
cobotic architecture (Figure 1), which allows a single actuator to deliver power to multiple degrees of freedom. Such
a scheme requires n+1 actuators for an n-degree-of-freedom
system: one large actuator to source power, and n small
actuators to modulate the transmissions for each degree of
freedom. The joints can still be operated independently, since
the transmissions are adjustable smoothly through an infinite
reduction, to both positive and negative ratios. The actuators
that modulate the transmissions for each degree of freedom
can be extremely small and low power, often one to two
orders of magnitude smaller than the single power actuator.
The transmissions draw power from a single common element actuator as needed, thus reducing the weight and power
requirements of the mechanism. Only one set of high power
electronics and drive-train components are needed.
IV. SAFETY
Joint 1
Joint 2
Joint n
v1
v2
vn
IVT 1
IVT 2
IVT n
Z
Common
Element
Fig. 1. Parallel power flow from a single large actuator is mediated by a
series of infinitely variable transmissions (IVTs), which modulate the power
flow to each joint.
III. ENERGY SAVINGS
Parallel cobotic architecture allows for significant electrical energy savings for a variety of reasons, all stemming
from the adjustable transmission ratio. If a high output speed
but low load is required, the transmission ratio is adjusted
to be a small reduction, and the apparent inertia of the
motor at the output is small. Thus little electrical power is
wasted to spin up and down the inertia of the motor and
drivetrain. Conversely, if a low output speed but high load
is required, the transmission ratio is adjusted to be a large
reduction, and very little motor torque need be applied, thus
minimizing resistive heating losses of the motor windings. In
fact, infinitely variable transmissions can be adjusted to an
infinite reduction, such that no motor torque is required to
hold large loads. The transmission itself can act as a clutch,
holding loads passively without the need for an additional
clutch.
The redundancy of the n + 1 actuator parallel cobotic
system is utilized to drive the system with the most power
efficient set of reduction ratios and common actuator speeds
at all times. The common element actuator is operated at
an efficient speed nearly all of the time. In addition, the
cobotic architecture allows for the ability to both clutch
or decouple joints without any additional actuators beyond
the low-power steering actuator for each IVT. With cobotic
architecture, no electrical power is expended to resist forces
in constrained directions. Electrical power is spent only to
provide effort along the current motion direction. Rolling
constraints in the transmission elements, not electrical power,
resist forces orthogonal to the current motion direction. Only
joints involved in the current motion direction draw off power
from the single common element actuator. Electrical power
is consumed only to do work on the inertia of limbs or the
outside world, not merely to sustain forces. An extensive
comparison of the power and energy efficiency of cobotic
versus conventional fixed ratio drivetrains is presented in [1].
Cobotic transmissions have a built in safety feature as well.
Since they rely on frictional contacts to transmit power, the
preload force at these contacts can be set to slip when a
certain output force or joint acceleration is exceeded. This is
beneficial to both the safety of the user, and for the drivetrain.
Impacts delivered to the limb merely result in slip in the
transmissions, since they are backdrivable up to the infinitely
variable component that utilizes friction contacts. The variable reduction ratio also allows lower actuator speeds, and
therefore lower kinetic energies accumulate in the actuators,
reducing the risk of damage to drivetrains during impacts.
Finally, the variable ratio drivetrain also allows for the use
of smaller actuators that are lighter and less powerful, and
therefore are less capable of injuring a user.
V. CONTROLLABILITY
The prosthetics community has seen the use of synergistic prehension and dual ratio drivetrains as alternatives
to infinitely variable transmissions for extending the forcevelocity regime in which a joint can be operated. Controller
implementation for either alternative is not trivial since synergistic prehension requires two actuators and the dual-stage
drivetrain requires a discrete switch between ratios. Typical
prosthetics also require clutches to hold high loads since the
passive lift rating of prosthetic limbs is typically much higher
than their active lift. However, these clutches hamper precise
control of impedance when engaging and disengaging. Cobot
transmissions exhibit this clutching ability without the need
for an additional clutch mechanism since they can act as a
clutch or brake when set to an infinite reduction. Conversely,
if the cobotic transmission is set to a zero reduction, the
output is effectively decoupled from the input, therefore
putting the mechanism in a passive, backdrivable mode.
The dynamic range of a cobot, with an adjustable reduction ratio, more closely matches the range of natural
impedances deliverable by human limbs than a fixed reduction system. Although cobots are controlled as admittance
devices, by allowing motions based on the applied force, they
do not suffer from the high inertia, friction and backlash that
normally exist in a highly geared admittance device. Cobots
excel at rendering the wide range of impedances that natural
human limbs are capable of. Also, the variable transmission
ratio allows for a low apparent inertia (small reduction ratio),
a requirement for simulating low impedances, but when
strength is required, large reduction ratios can be chosen.
Cobotic devices control the relative velocities of their
joints by modulating IVTs with small steering actuators,
thus directing the single instantaneous motion freedom characteristic of cobots, regardless of the dimension of their
configuration space. The parallel cobotic system with n + 1
actuators has n adjustable velocity constraints in the transmissions, yielding a single degree of freedom system. The
dynamics along this single motion freedom are controlled via
a single power injector, or by a human operator in the case
of a passive cobot. Rolling constraints in the transmission
elements, not electrical power, resist forces orthogonal to
the current motion freedom. This leads to a natural stability
when rendering virtual constraints, since the instabilities
that plague conventional haptic displays, which arise from
sampling in discrete time and space and exciting structural
resonances, cannot impact any control loops in the constraint
directions.
VI. CONCLUSION
We propose cobotic technology as an infinitely variable
transmission architecture that when applied to prosthetics
will yield reductions in weight, energy consumption, as
well as improve safety and controllability. Cobots allow
for variable back-drivability, high power efficiency, precise
control of output force and velocity at low output speeds,
and a single power actuator for multiple degrees of freedom
without the need for brakes or clutches. We have previously demonstrated the scalability of cobotic technologies
to produce high degree-of-freedom, high bandwidth haptic
devices [1], and intend to employ technology gained from
these devices in the field of prosthetics.
R EFERENCES
[1] E.L. Faulring. The cobotic hand controller: Design, control and analysis
of a novel haptic display. Ph.D. Dissertation, Northwestern University,
2005.
[2] R.B. Gillespie, J.E. Colgate, and M.A. Peshkin. A general framework
for cobot control. IEEE Transactions on Robotics and Automation,
17(4):391–401, 2001.
[3] C.A. Moore. Design, construction, and control of a 3-revolute arm
cobot. Ph.D. Dissertation, Northwestern University, 2001.
[4] M.A. Peshkin, J.E. Colgate, W. Wannasuphoprasit, C.A. Moore, R.B.
Gillespie, and P. Akella. Cobot architecture. IEEE Transactions on
Robotics and Automation, 17(4):377–390, 2001.
[5] W. Wannasuphoprasit, R.B. Gillespie, J.E. Colgate, and M.A. Peshkin.
Cobot control. In International Conference on Robotics and Automation, pages 3571–3576, Albuquerque, NM, 1997.
[6] T. Worsnopp, M. Peshkin, J.E. Colgate, and K. Lynch. Controlling
the apparent inertia of passive human-interactive robots. In IEEE
International Conference on Robotics and Automation, pages 1179–
1184, New Orleans, LA.