IEEE Robotics & Automation Magazine - March 2013 - 23

Position Control
Besides safe interaction, maintaining accurate position control is also important to assistive robotics for object grasping,
manipulation, and autonomy. Position control is very effectively used and optimized in industrial applications, e.g., in
the automobile industry. Typically, these robots have highly
stiff structures driven by stiff joints, and their eigenfrequencies are implicitly high; so, the control bandwidth can be large.
For a fast position control, a lower mass allowing higher
accelerations is beneficial, but it increases the disturbance
sensitivity of the system and puts higher requirements on
control. Therefore, system mass is often a tradeoff among the
required speed, available power, and control constraints.
High-resolution joint encoders provide accurate feedback,
and therefore, repeatability.
Examples of systems that ensure an accurate position
control and repeatability by stiff structures/joints and
high-resolution joint encoders are [3], [10]-[14], [16],
[18]-[20], and [25]-[29].
Energy
The energy criterion relates to the capability of storing energy
and reusing it at a later stage and the possibility of using and
shaping intrinsic system dynamics to achieve a desirable
motion with little actuation, which is energetically efficient.
Being energy efficient is an important feature, since assistive
robotics are preferably mobile, and therefore should be able to
operate independently from an external power source for a
considerable amount of time.
Storing energy can only be done by incorporating a physical storage element, like a mechanical mass or spring, or an
analog element in an arbitrary domain, e.g., an electrical
inductor or a capacitor. A moving mechanical mass stores
kinetic energy, so if the kinetic energy at the load of a robotic
arm is transferred to this mass, the energy can be stored and
reused later (e.g., regenerative braking). Since storing much
energy in this way requires a large mass and high speed, a
mechanical elastic element may be used instead. The kinetic
energy is then stored as potential energy by, for instance,
using the mechanical springs. Examples of systems that have
the capability to store energy using mechanical springs are
DLR HASy, Robonaut 2, and MIA Arm.
Useful intrinsic system dynamics are any dynamics that
require no additional external energy that can contribute
to a certain desired motion or behavior, and are most
often oscillatory of nature. Adjustment of these intrinsic
dynamics offers flexibility in achieving energy-efficient
motion, i.e., a desired motion that is primarily the consequence of intrinsic dynamics as opposed to actuation. The
MIA Arm and the DLR HASy can tune intrinsic system
dynamics by tuning their joint stiffness and, thereby, possibly adjusting useful resonance frequencies.
Adaptability
The adaptability criterion of an assistive robotic system
relates to its ability to change system properties to provide a

dynamic tradeoff between safety and performance. This
means that, for instance, safety can be decreased if more
performance is needed (e.g., a higher speed has to be
reached), which may be necessary due to unknown arbitrary environments and conditions. Just like interaction
safety, adaptability can be achieved by either active measures, e.g., shaping the performance limits and using the
active impedance control, or by passive measures e.g., variable stiffness joints.
Adjusting motor current limits to reduce the maximum
possible force that can be applied by the joints (and therefore
the arm's end-effector), or adjusting the speed and acceleration limits increases safety by reducing the possible force with
which a human may be hit. An example of a system that uses
adjustable performance (speed) limits is iARM.
The active impedance control allows the force that is
applied to a load to be controlled. Since this is done actively,
e.g., in software, the virtual elastic element can be tuned to a
desired value required for safety, performance, or other conditions. Examples of systems that have shown ability to adapt
the safety and performance tradeoff by means of active
impedance controls are KARES II, WAM Arm, Elumotion
RT2, DLR LWR-III, and Modular Prosthetic Limb.
The variable stiffness joints can adapt the physical stiffness
perceived at the output side of a joint. Decreasing this physical
stiffness means the joint is perceived as a softer spring, while
increasing stiffness means that the joint is perceived as a
stiffer spring. The former implies an intrinsic increase in the
safety and probable degradation of the performance, whereas
the latter implies an intrinsic decrease in the safety but an
improved performance. Examples of systems with variable
stiffness joints are DLR HASy and MIA Arm.
Discussion
Based on usage surveys, general usage scenarios have been
reported in the section on the robotic arms. From these preferences, we have proposed five criteria to evaluate the assistive
robotic arms. Several possible implementations to meet the
five criteria are discussed in the previous sections and examples of current assistive robotic systems are given that reflect
these implementations. Some of the criteria are not completely independent and are related or coupled with another
criterion. The energy criterion, for instance, is related to interaction safety and shock robustness if the energy-storing capabilities are achieved by implementing a physical elastic
element, interaction safety is achieved by robust force control,
and shock robustness is achieved by the capability to store
kinetic impact energy as potential energy in the elastic element. However, the reverse does not necessarily hold; if shock
robustness is provided by slip clutches, no energy-storing
capabilities are implicitly provided. The same holds for adaptability and interaction safety; a suitable system adaptability,
according to the criterion, means good interaction safety as a
consequence, but suitable interaction safety may be achieved
by static implementations (for instance, by using the lowpower motors), which results in low adaptability.
March 2013

IEEE ROBOTICS & AUTOMATION MAGAZINE

Table of Contents for the Digital Edition of IEEE Robotics & Automation Magazine - March 2013