• DOI:10.1101/SQB.1990.055.01.074
• Corpus ID: 18818376

Understanding sensorimotor feedback through optimal control.

@article{Loeb1990UnderstandingSF,
  title={Understanding sensorimotor feedback through optimal control.},
  author={Gerald E. Loeb and William S. Levine and J He},
  journal={Cold Spring Harbor symposia on quantitative biology},
  year={1990},
  volume={55},
  pages={
          791-803
        }
}

• G. Loeb, W. Levine, J. He
• Published 1990
• Biology
• Cold Spring Harbor symposia on quantitative biology

The general problem of "control" may be divided into two main categories, often designated open-loop and closed-loop (see Fig. 1). We adopt here the convention of using "controller" to designate a device that formulates a set of commands intended to change the state of a system, which commands are executed openloop, i.e., without modification during the task. (We emphasize that this is a shorthand terminology, not to be confused with the standard usage in control theory denoting both openand closed-loop components.) We use the term "regulator" to designate a device that attempts to stabilize the state of a system, generating only closed-loop commands, i.e., responses to deviations from the state detected by sensors in the system (Bryson and H o 1975). Obviously, most complex systems require a mixture of the two types of control. In sensorimotor neurophysiology, it is common to depict these kinds of controls schematically as if they were separate, even though the anatomical and physiological features of the structures in which they reside suggest a close interrelatedness. For example, the control of locomotion in quadrupeds has been seen as divided between a spinal central pattern generator (CPG) , which generates an open-loop program of muscle activation, and various reflexes, also largely spinal, which adjust the activation in response to internal errors and external perturbations. Liddell and Sherrington (1925) first proposed the concept of the motoneuron as the "final common path" whereby the various sources of control signals would be summed to result in the net command to a motor unit. There is now little doubt that the motoneuron, like most central neurons, serves to integrate many disparate sources of input into a single-dimensional output. However, the later notion that these motoneuronal inputs are distinguishable into open-loop and closed-loop types deserves reexamination. Ironically, Sherrington (1910) himself felt that locomotion was produced primarily through reflex pathways, i.e., from the sequential combination of sensory signals arising from the musculoskeletal mechanics of the limb. T h e need for an open-loop controller of locomotion was firmly established later by the surprisingly natural temporal patterns of motoneuronal output reported during fictive locomotion in paralyzed, de-