Motor Control, Biological and Theoretical

6 Biological motor control can be characterized as a problem of con7 trolling nonlinear, unreliable systems whose states are monitored 8 with slow and sometimes low-quality sensors. In response to 9 changing sensory inputs, internal goals, or motor errors, the motor 10 system must solve several basic problems: selection of an appro11 priate action and transformation of control signals from sensory to 12 motor coordinate frameworks; coordination of the selected move13 ment with other ongoing behaviors and with postural reflexes; and 14 monitoring the movement to ensure its accuracy. These stages may 15 be interlinked, so that separation of any one particular problem into 16 these individual stages may not be possible. This article describes 17 some of the ways we think that biological motor systems solve 18 these tasks, based on principles (and terminology) whose origins 19 are in engineering and cybernetics. The field of cybernetics has 20 developed from Norbert Wiener’s initial ideas on communication 21 and control theory in complex mechanical and biological systems, 22 which focused on feedback mechanisms. 23 A motor control system acts by sending motor commands to a 24 controlled object, often called the “plant,” which in turn acts on its 25 local environment (Figure 1). The plant or the environment has one 26 or more variables that the motor system attempts to regulate, either 27 to maintain them at a steady reference level in the face of distur28 bances (a “regulator”) or to follow some changing reference value 29 (a “controller”). The motor control system may make use of sen30 sory signals from the environment, from its reference inputs, and 31 from the plant to determine what actions are required. Sensory 32 inputs from the plant can provide information about the state of 33 the controlled object. Here, the state can be considered as all rele34 vant variables that adequately describe the controlled object. But 35 note that the sensory inputs to the controller do not necessarily 36 provide direct measures of the true state of the system: They may 37 be inaccurate or delayed, as discussed later. If controller output is 38 based on signals that are unaffected by the plant output, it is said 39 to be a feedforward controller: The feedforward control path is the 40 thick line from left to right in Figure 1A1, which requires no return 41 signals. If the controller output is instead based on a comparison 42 between the reference and the controlled variables, it is a feedback 43 controller (Figure 1B1): The control pathway is a closed loop. One 44 can add more complex control strategies to these simple systems 45 (Figure 1A2, 1B2), as described in more detail below. 46 The advantage of feedforward control is that it can, in the ideal 47 case, give perfect performance with no error between the reference 48 and the controlled variable. The main disadvantages for biological 49 systems include the potential difficulty in generating an accurate 50 controller for a complex system and the lack of error corrections. 51 If the controller is not accurate, if the plant is unreliable, or if 52 unexpected external disturbances occur, output errors go un53 checked. Since no biological system can be both perfectly accurate 54 and perfectly free of external disturbances, error correction is usu55 ally necessary. In contrast, the major advantage of negative feed56 back control lies in its very simple, robust strategy. The controller 57 drives the plant so as to cancel the feedback error signaled by the 58 comparator. Because it constantly seeks to cancel the error, it op59 erates well, even without exact knowledge of the controlled object 60 and despite internal or external disturbances. But feedback control 61 strategies also have disadvantages: Errors cannot be avoided but 62 must occur and be corrected, and feedback control—especially in 63 biological systems—tends to be slow.