The synergy of human arm and wearable robot systems studies the compatibility of a wearable robot system and its operators. Ideally, the motion of a wearable robot system should be dynamically transparent to a healthy operator, sensitively responsible to the voluntary and involuntary motions of its operator. When used for robot-assisted stroke rehabilitation and/or power augmentation, a wearable robot system is expected to provide assist to the operator's motor skills and correct the abnormal movements resulted from motor disabilities. Inspired by the control strategies of human motor system, the robot systems can be more dexterous in manipulations and more adaptable to various tasks and uncertain environments.
The study on synergy of human arm and wearable robot systems intends to find out the motor control strategies that dominate human arm movements, in order to apply to the control of robotic systems, such as the upper limb exoskeletons and the surgical robot systems. This research experimentally investigated the arm movements in various tasks, to enhance the understanding of human motor control and to benefit the control robotic systems. With the data collected from point-to-point reaching movements, it proposed a redundancy resolution methodology based on multi-criterion performance optimization, which provides real-time predictions of arm postures with improved accuracy. Based on the comparison between reaching and reach-to-grasp movements, it further studied arm postures affected by grasping orientations and pointed out the control strategy that dominates the macro- and micro- structure coordination. From both healthy subjects and stroke patients, experimental data of unimanual and bimanual movements are recorded for identical tasks. The statistical analysis shows that motor function of the paretic arm can be improved when moving symmetrically with the non-paretic arm, which reveals the beneficial effects of the inter-arm coupling on stroke rehabilitation.
The dissertation contributed to answer the following three research questions:
* How to predict the natural arm postures of reaching movement?
The kinematic redundancy of the human arm enables the elbow position to rotate about the axis going through the shoulder and wrist, which results in infinite possible arm postures when the arm reaches to a target in a 3-dimensional workspace. To infer the control strategy the human motor system uses to resolve redundancy in reaching movements, this research compares five redundancy resolution criteria and evaluates their arm posture prediction performance using data on healthy human motion. Two synthesized criteria are developed to provide better real-time arm posture prediction than the five individual criteria. Of these two, the criterion synthesized using an exponential method predicts the arm posture more accurately than that using a least squares approach, and therefore is preferable for inferring the contributions of the individual criteria to motor control during reaching movements. As a methodology contribution, this paper proposes a framework to compare and evaluate redundancy resolution criteria for arm motion control. A cluster analysis which associates criterion contributions with regions of the workspace provides a guideline for designing a real-time motion control system applicable to upper-limb exoskeletons for stroke rehabilitation.
* How do the arm joints coordinate in reach-to-grasp movement?
Reach-to-grasp movements are widely observed in activities of daily living, particularly in tool manipulations. In order to reduce the complexity in redundancy resolution and facilitate upper-limb exoskeleton control in reach-to-grasp tasks, we studied joint coordination in the human arm during such movements. Experimental data were collected on reach-to-grasp movements in a 3-dimensional (3D) workspace for cylinder targets of different positions and grasping orientations. For comparison, reaching movements toward the same targets are also recorded. In the kinematic analysis, the redundant degree of freedom in human arm is represented by the swivel angle. The four grasping-relevant degrees of freedom (GR-DOFs), including the swivel angle and the three wrist joints, behave differently in reach-to-grasp movements comparing to how they behave in reaching movements. The ratio of active motion range (R-AMR) is proposed for quantitatively comparison the task-relevance of the GR-DOFs. Analysis on the R-AMR values shows that the task-relevant GR-DOFs are more actively used, while the task-irrelevant joints are left uncontrolled and maintain their neutral positions. Among the task-relevant GR-DOFs, the smaller joints (micro-structure) are more actively used than the larger joints (macro-structure). The coordination of the task-relevant GR-DOFs is shown to be synergistic. Analysis of the acceleration/deceleration at the GR-DOFs indicates different levels of voluntary control in three phases of the movements. The study of the characteristics of the joint coordination in reach-to-grasp movements provides guidelines for simplifying the control of the upper limb exoskeleton.
* Can the bilateral training benefit the post-stroke stroke recovery?
Bilateral training has been used for post-stroke rehabilitation of patients with paretic upper arms. This training method exploits the inter-arm coupling in bimanual movements which tends to synchronize the paretic arm with the non-paretic arm for motor function recovery. Different levels of motor function recovery have been reported in clinical studies, yet the efficacy of bilateral training is still not clear. As a result, this paper collected data on bimanual symmetric reaching movements in a 3-dimensional (3D) workspace from ten healthy subjects and eight chronic stroke patients, to investigate the effects of inter-arm coupling. In reaching experiments, identical tasks are also performed unimanually and recorded. Mapping variables are proposed to measure arm behavior changes and to quantify the similarity of the movements of the two arms. Among the four proposed mapping variables, mapping linearity and mapping complexity were found to reflect both the motion complexity and the task-relevance of a joint, while the other two variables captured changes in motion range and relative velocity. Statistical analysis shows that inter-arm coupling increases inter-arm symmetry more at task-relevant joints than at task-irrelevant joints. To achieve symmetric bimanual movement, the dominant arm of healthy subjects and the non-paretic arm of stroke patients deviate more from their unilateral behavior than the non-dominant/paretic arms. For stroke patients, the effect of inter-arm coupling on arm function depends on the task-relevance of a joint. During bimanual movement, joint behavior at task-irrelevant joints is different from behavior during unimanual movement, which implies that bimanual rehabilitation may help break abnormal movement patterns for these joints. For task-relevant joints, the average arm posture of the paretic arm is significantly higher in bimanual movement, which indicates the possible efficacy of bilateral rehabilitation.