Investigation of control of human balance-recovery reactions


Tezin Türü: Doktora

Tezin Yürütüldüğü Kurum: Orta Doğu Teknik Üniversitesi, Fen Bilimleri Enstitüsü, Fen Bilimleri Enstitüsü, Türkiye

Tezin Onay Tarihi: 2015

Öğrenci: NURDAN BİLGİN

Danışman: MUSTAFA KEMAL ÖZGÖREN

Özet:

In this study, a natural control law has been conjectured, which is assumed to be applied by the central nervous system of a human being in order to maintain the erect posture after being exposed to a suddenly occurring impulsive disturbance. The control law is conjectured as PD (proportional-derivative) control, because it has to be compatible with the physiological facts. The conjectured control law has been validated by comparing a set of experimental results with the corresponding set of simulation results generated by applying the conjectured control law to a biomechanical model of a typical human being. For the sake of simplicity, the model is confined to the sagittal plane with three degrees of freedom having the ankle, knee, and hip joints as the only actuated joints. The other joints are assumed to be kept fixed. The experiments are carried out in a way compatible with the model by using a tilt platform and by requesting the human subjects to keep their hands in their pockets while standing on the platform. During the experiments, the impulsive disturbance is given by tilting the platform suddenly and then the balance recovery motions of the human subject are recorded. The experiments and simulations are repeated several times on different human subjects. In all the test runs, it has been possible to match the simulation and experimental results only if the simulations are made by changing the gains of the conjectured PD control law as particular functions of time. The experimentally inferred fact of changing PD control gains leads to a major hypothesis that the central nervous system applies the conjectured PD control law by changing its gains according to a certain adaptation law. Naturally, it is difficult to estimate this adaptation law because it is probably very much dependent on the past physiological and psychological experiences of the human beings. Nevertheless, it may be possible to estimate at least a set of major arguments along with a functional relationship between them and the adapted gains. In this study, it has been possible to arrive at such an estimation by means of the renown canonical correlation method. When this method is applied to correlate the experimental and simulation results, it has been found out that the adaptation law seems to be a linear relationship that gives the PD control gains in terms of the error state and input variables as long as they remain small in magnitude. Furthermore, upon examining the experimental results gathered from the repeated runs for different human subjects, another evident correlation is detected, which is between the initial posture of the human subject on the platform and the subsequent pattern of the balance-recovery response that occurs after a sudden tilt of the platform. In order to express this correlation mathematically, the initial postures and the balance-recovery responses are classified by using classification algorithms. Afterwards, the expression obtained for this correlation has been tested and verified to a large extent with Monte Carlo simulations by using the decision tree created in accordance with the classification.