The use of linear multi-input/multi-output control theory to study posture control

Summary form only given. It is argued that recent results in linear multi-input/multi-output (MIMO) control theory provide new insights into the control of posture and that these methods can provide even more insight once they begin to motivate experiments. The first step in applying linear MIMO control theories to the control of posture is to create a mathematical model of the dynamic system to be controlled. Such a model, in the absence of simplifying assumptions, would be nonlinear and of very high state dimension. The size of the problem can be dealt with by focusing on a particular aspect that is fairly independent, such as the control of the posture of one leg. It is then at least plausible that the effect of the neglected portions of the system is only to perturb the controller for the leg. The next step is to model the sensors and to determine which states are available to the controller. A linear state-feedback control theory based on the linear quadratic regular (LQR) is discussed. This theory produces linear state-feedback controls that minimize a quadratic performance measure. The performance measure has some designer specified parameters that can be used to vary the relative importance of control versus state errors. This approach has been used to derive optimal linear state-feedback controls for regulating the posture of the 36 state limb model