Elderly individuals and those affected by neuromuscular disorders are frequently not able to independently maintain seated balance. As a result, these individuals are limited in performing activities of daily living, and are susceptible to an increased risk of falling and secondary health complications. To develop therapies and targeted interventions for seated instability, it is essential, however, to first quantify the mechanisms responsible for controlling seated balance. In this context, classical system identification techniques are a promising tool for obtaining a quantitative description of such mechanisms. Motivated by these considerations, the objective of this M.Sc. research project was to quantify, using advanced system identification techniques, the active and passive control mechanisms, the muscular dynamics, and the sensorimotor time delay in seated balance control of non-disabled individuals.

14 young, non-disabled individuals were perturbed while sitting using mild, mechanical surface perturbations. The body kinematics, muscle activity, and ground reaction forces were recorded during the perturbations. Neuromusculoskeletal time series, including the body sway, the joint torque using top-down and bottom-up inverse dynamics, and the weighted electromyography representing neural activation, were calculated. Using the joint input-output system identification technique, non-parametric estimates of the active control components (neural dynamics and sensorimotor time delay) and of the active-passive control components (neural dynamics, mechanical dynamics, sensorimotor time delay, and muscular dynamics) were obtained. Parametric estimates of these components were computed using model fitting. The parameters’ accuracy was estimated using goodness-of-fit (GOF), the Akaike information criteria (AIC), and variance-accounted-for (VAF). The stability of the identified models was then assessed via a pole-zero analysis of the characteristic equation. Finally, the identified models were implemented in simulations to assess the robustness of the model parameters.

While the thesis presents results for both the top-down and bottom-up inverse dynamics, only the top-down results are described in this abstract unless stated otherwise. For both the active and active-passive controller components, the frequency response was approximately constant for lower frequencies (<​ 0.4 Hz) and then steadily rose as the frequency increased. The active control component’s frequency response had a phase of 30 degrees for lower frequencies and steadily rose as the frequency increased;​ however, it saturated around 110 degrees as the frequency reached approximately 3 Hz. The active-passive control component’s frequency response had a constant phase of approximately 180 degrees for the lower frequencies (<​ 1Hz) that gradually increased to approximately 185 degrees at 2.5 Hz and then settled at approximately 180 degrees. The across-participant variability of the non-parametric estimates of the active and active-passive control components was small. The neural dynamics were identified as a proportional-derivative (PD) controller with acceleration feedback;​ the sensorimotor time delay as an exponential decay function;​ the mechanical dynamics as a PD controller;​ and the muscular dynamics as a second-order transfer function. The fitting of the active control components using the stated models provided GOF, AIC, and VAF ranges of 99.2–99.8%, 1.1–1.5, and 29.0–60.8%, respectively. Similarly, the fitting of the active-passive control components using the stated models provided GOF, AIC, and VAF ranges of 99.9–99.9%, 0.4–0.5, and 97.7–99.7%, respectively. The stability analysis identified the neural dynamics, the sensorimotor time delay, and the mechanical dynamics (using bottom-up inverse dynamics) to produce a stable characteristic equation. The difference between the experimental and simulated parameters was low in most cases.

In this study, the mechanisms of seated balance have been successfully quantified for non-disabled individuals. The gained insights support the notion that closed-loop feedback control contributes to stabilizing the upper body during sitting and that a velocity-acceleration-based strategy is utilized for active control. The identified parameters can furthermore be used as a normative benchmark for quantitatively and mechanistically assessing the severity of seated imbalance in affected individuals, with the goal of optimizing rehabilitation therapies and interventions.