IEEE Robotics & Automation Magazine - March 2013 - 22

The interaction safety of robotic arms is a measure for
the suitability of human-robot interaction and is associated with the intrinsic and extrinsic, i.e., the passive and
active, safety measures. The intrinsic safety measures refer
to measures that increase the safety even when no power
is supplied to the system, while the extrinsic safety measures refer to the safety measures that rely on external
power. Possibilities for achieving interaction safety that
have been shown in the literature include limiting an
arm's performance, incorporating joint backdrivability,
using an active impedance control, and implementing
intrinsically compliant joints.
Incorporating mechanical or electrical constraints in the
system can lead to a higher level of safety. If, for instance, the
maximum speed and force of the arm are kept low and the
arm collides with a human, there is little chance of injury.
Such constraints on the mechanical or the electrical characteristics are defined here
as the performance limits.
Limiting the performance
The position control is
of a robotic arm can be
done by, for instance, limvery effectively used and
iting the arm's movement
speed and acceleration in
optimized in industrial
space, the end-effector
force, and the maximum
applications.
possible payload. The
speed limits in combination with low arm inertia ensure that the momentum with
which a human is accidentally hit is kept low. Examples of
such systems are JACO [10] and iARM [11]. Contrary to the
current limits that can be changed on demand, using lowpower motors is a way of statically and intrinsically limiting
an arm's acceleration, force, and payload, as done in Weston
[12]. Also, Bridgit [13] and RAPUDA [14] are systems that
limit their performance to increase safety. The iARM is also
an example of a system that incorporates slip clutches, i.e.,
mechanical power transmission couplings between the actuator and the joint. They disconnect or slip the transmission
when the applied joint torque exceeds a certain maximum,
thus limiting the end-effector payload and the possible force
on the environment [15].
Using backdrivable joints is another way of increasing
safety. Backdrivable joints do not resist an external output
motion, depending on the amount of current supplied to (and
therefore torque supplied by) the joint motors. Therefore, a
user is able to manipulate the system externally without
feeling a rigid arm structure. Examples of systems using
backdrivable joints are JACO and WAM Arm [16].
The active impedance control [17], often applied to intrinsically stiff joints that have a stiff coupling between an actuator and joint, is a well-known strategy of incorporating a
means of safety in a robotic system. As opposed to the position control where the error between a desired and actual
position has to be minimized, the impedance control regulates the interaction between an end-effector and its environ22

IEEE ROBOTICS & AUTOMATION MAGAZINE

March 2013

ment. This way, a virtual compliance can be included at the
joint level between the motor and the output, causing the
human operator to perceive a softer arm. Because of this soft
behavior, the arm becomes safer to use. Examples of systems
that use this type of control are KARES II, WAM Arm,
Elumotion RT2 [18], DLR LWR-III [19], Modular Prosthetic
Limb [20], and DLR HASy [21].
In intrinsically compliant joints, compliance or stiffness
during the interaction between the end-effector and the
environment is realized by a physical elastic element, e.g., a
mechanical spring which is put between the actuator and
the joint, as opposed to a virtual elastic element in the case
of active impedance control. The stiffness felt during interaction is physical stiffness due to the elastic element. Therefore, these joints are intrinsically safe during interaction,
since they do not rely on limited bandwidth controllers or
possibly unreliable measurements. An example of such a
system is Robonaut 2 [22], which uses a series of elastic
joints to actuate the arm. Much like the active impedance
control, which mimics elastic elements of varying compliance, variable compliant joints or variable stiffness joints
[23] can adjust their physical stiffness between the actuator
and the joint. In this type of joint, at least two motors are
used to simultaneously control position and the stiffness.
Examples of such systems are DLR HASy and MIA Arm
[24]. Although the active impedance control can still be
applied to intrinsically compliant joints, safe interaction
behavior is guaranteed by the design.
Shock Robustness
Shock robustness is a criterion for estimating the amount of
damage an arm suffers upon external high-impact shocks.
Arm joints are the most sensitive to these shocks and therefore should be protected. Shock robustness can be achieved
by using a system bandwidth capable of absorbing the
shocks or by allowing the temporary mechanical decoupling
of the actuator from the joint. Both measures can prevent
damage to the motors and gearboxes upon the shocks.
A system bandwidth capable of reacting to the arbitrary
shocks can be realized only physically, i.e., by using the intrinsically compliant joints. Again, these joints contain a physical
elastic element between the actuator and joint, thereby converting kinetic impact energy to potential energy, as in DLR
HASy, Robonaut 2, and MIA Arm.
The mechanical decoupling of the actuator from the joint
upon high impacts prevents damage to the motors and gearboxes. This can be achieved by using the joint slip clutches,
which decouple the power transmission, thereby allowing the
joint to move passively and allowing the conversion of kinetic
impact energy to kinetic energy in the decoupled arm. Also,
the backdrivable direct-drive joints, i.e., the joints without
nonbackdrivable gearboxes, ensure that no damage occurs
since the kinetic impact energy does not have to be absorbed
by the gearbox, motor, or arm structure. Examples of systems
with slip clutches or backdrivable joints are JACO, iARM, and
WAM Arm.

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