Feedback and Feedforward

There is consensus that neither purely feedforward, open-loop control nor purely feedback control is adequate for the regulation of movement in a biological system. Feedforward control alone cannot work because it requires that all the parameters of the system (its inertia, the starting posture, muscle mechanics, etc.) be known with a precision that appears to be beyond the realm of biological sensors. Similarly, pure feedback control is inadequate because the feedback gain is low. If the feedback gain were higher, the inherent time delays would lead to instability.

Thus, movement regulation appears to involve a hybrid of feedforward and feedback. In this conceptualization (Fig. 3), the feedforward pathway involves a model of the inverse dynamics of the limb that is a prediction of the forces that would be required to generate a desired movement. Thus, the inverse model transforms desired kinematics into motor commands (kinetics) acting on the musculoskeletal system to generate movement (kinematics). If the inverse model is accurate, the actual kinematics will match the desired kinematics. As stated previously, the inverse model will never be completely accurate, and accuracy is achieved by means of feedback.

In Fig. 3, the sensory feedback signals the actual kinematics resulting from the motor commands. These are compared to the expected kinematics, obtained using an efference copy of the motor commands. This comparison requires a transformation from the kinetic signal carried by the efference copy to kinematics. Because the transformation is in the forward, causal direction from kinetics to kinematics, it is called a forward model. The feedback regulation would then provide for a comparison of the expected motion, derived from the forward model, with the actual motion that is sensed by various afferents. This error signal could be used as an input to the inverse model to update the motor commands.

In this scheme, a mismatch between the predicted motion and the actual motion would have two effects, occurring on two different timescales. First, the motor commands would be modified, through the inverse dynamic model, to correct the ongoing motion. Second, if errors persist over many trials, the inverse dynamic and the forward dynamic models would be modified so as to bring their predictions into better accord with the actual performance.

Note that this hybrid scheme bears some resemblance to the robotically inspired scheme for movement planning outlined previously. Recall that this scheme involved kinematic stages in which ultimately the desired trajectory was specified in terms of the motion of each of the joints. A second stage converted this desired trajectory into the torques at each of the joints (and forces at each of the muscles) required to produce the requested motion. In terms of the hybrid model just described, this stage corresponds to the feedforward, inverse dynamics model part of the process. What has now been added is a feedback controller, correcting for errors inherent in the feedforward stage.

Breaking Bulimia

Breaking Bulimia

We have all been there: turning to the refrigerator if feeling lonely or bored or indulging in seconds or thirds if strained. But if you suffer from bulimia, the from time to time urge to overeat is more like an obsession.

Get My Free Ebook

Post a comment