Modelling Motor Control Paradigms - Computational Neuroscience: A Comprehensive Approach - page 535

Figure 17.2
Logical decomposition of motor control into cascaded computational processes.

Figure 17.3
Internal and external forces in a kinematic chain. In the usual control paradigm F ext
is related to the load and u i 's are the controlled actuator forces. In the passive motion
paradigm F ext is related to the planned trajectory and u i 's correspond to the passive
compliance of the joints to the planned pull.

5. Actuator control: the typical robotic actuators are force-controlled and thus the
output of the inverse dynamics calculations can be directly fed to the control
variables of the actuators.

17.2.2 The Achilles' heel of feedback control

In practice, the scheme of Figure 17.2 cannot be applied directly because the dy-
namical model is uncertain and partially unknown. Thus, on top of a computational
process structured as an approximation of Figure 17.2 usually there is a layer of feed-
back control which can use positional and/or force feedback ( Figure 17.5 ). The main
problem of feedback control, particularly from the perspective of using it as a possi-
ble paradigm for modelling biological motor control, is that it is critically dependent