Page 200 - Handbook of Biomechatronics

Upper-Limb Prosthetic Devices 197

(bidirectional interface with the brain), and (b) the low-level, which ensures
that these commands are followed with the required accuracy.
Although actuation adds the capability for motion and force application,
a low-level control system is needed to interpret feedback signals and send
appropriate currents to motors so as to achieve the desired motion, or apply
the desired force to the environment. In a sense, the control system is
required to recreate a force-velocity relationship for the prosthesis similar
to that of the missing limb; then the prosthesis is “transparent” and felt as
an extension of the patient’s body.
To develop a control system, one needs to have some knowledge of the
dynamics of the system to be controlled. As this has many-DoF and is
described by nonlinear equations of motion, the control system should be
nonlinear, too, making its design more involved that for single-DoF linear
systems. Prostheses have much in common with robotic arms and exoskel-
etons (Proietti et al., 2016), therefore control strategies that apply to them
can be classified as position control, force control, and interaction control.
In position control, the aim is to ensure that all controlled variables, usu-
ally joint angles or displacements, achieve their commanded value at the
right time. This must be achieved in the presence of friction, gravity, and
even external disturbances, such as loads or impacts, and requires position
sensors such as encoders or potentiometers. A large number of control
schemes exist such as PID, model-based, LQR, adaptive, etc., (Bhuiyan
et al., 2015). However, this type of control is only appropriate when the
limb moves in free space.
In force control, the aim is to have the limb apply a desired force or
moment to the environment, as for example, when using a hand tool, such
as a drill. This type of control requires a force sensor, so that the applied force
is available for feedback reasons and appropriate correction by the controller.
As it is not possible to control simultaneously the velocity and the force of
any body, position, and force controllers are incompatible. To have them
both, switching between the two modes would be required. Even then, a
delay in switching from position to force control can have undesired effects
on the prosthetic and the patient itself.
This problem is addressed by impedance control (Hogan, 1985) and its
variants (Calanca et al., 2016), where the aim is to control the relationship
between velocity and force, without any switching of controllers. Although
this is quite appealing for prostheses applications, the tuning of the control
gains is not easy and usually it is task dependent. Recently Variable imped-
ance actuators were introduced, where the controller takes into account the