UC Santa Cruz

The human arm including the shoulder, elbow, wrist joints and exclusion scapular motion has seven degrees of freedom(DOF) while positioning the wrist in space and orientating the palm is a task that requires six DOF. Given the redundant nature of the arm which has one more DOF than is needed to complete the task, multiple arm configurations can be used to complete a task based on none unique solution for the inverse kinematics. Despite this mathematical difficulty, the human motor control provides an unique solution for the arm redundancy as the arm moves in space. Resolving this redundancy is becoming critical as the human interacts with a wearable robotic system(exoskeleton) which includes the same redundancy as the human arm. Therefore, the inverse kinematics solution resolving the redundancy of these two coupled systems must be identical in order to guarantee a seamless integration.

Creating a proper control scheme between a wearable robot and human arm starts from an understanding of the redundant nature of the human arm. The redundancy of the arm can be formulated kinematically by defining the swivel angle - the rotation angle of the plane including the upper and lower arm around a virtual axis connecting the shoulder and wrist joints fixed in space. Then a global exoskeleton robot control scheme targeted for the natural human robot interaction will be achieved by providing a robot with the precise swivel angle estimation for the given kinematic and dynamic states of the human arm. In order for this, we first study human motor control mechanism for the simple reaching and grasping tasks from a kinematic point of view. Analyzing reaching tasks recorded with a motion capture system indicates that the swivel angle, which defines the redundancy of the human arm, is selected such that when the elbow joint is flexed, the palm moves toward the head for any wrist position. Based on these experimental results, a new criterion to resolve the human arm redundancy is formed and this criterion is to maximize the projection of the longest principle axis of the manipulability ellipsoid for the human arm on the vector connecting the wrist and the virtual target on the head region. For more realistic and natural human arm movement, we additionally considered the redundancy based on the dynamic criterion which minimizes the mechanical work done in the joint space for each two consecutive points along the task space trajectory. The the swivel angles from the kinematic and dynamic criteria were linearly combined with different weight factors for the unified the swivel angle. Post processing of experimental data collected with a motion capturing system indicated that by using the proposed synthesis of redundancy resolution criteria, the error between the predicted swivel angle and the actual swivel angle adopted by the motor control system was less then five degrees. This result outperformed the prediction based on a single criteria and showed that the kinematic constraint is dominant in a simple reaching and grasping tasks that frequently occurs in our daily life. In order to define the redundancy resolution mechanism for more generalized human arm movement, the effect of the wrist orientation on the redundancy of the human arm was superimposed onto the wrist position based swivel angle estimation. By applying the above inverse kinematics mechanism mimicking the natural human arm movement to the wearable robot, wearer can expect the synchronized movement with robot for unconstrained natural human arm movements. Finally, to accommodate the unnatural movement pattern such as avoiding obstacle, purely reactive task space admittance control based on multiple force sensors is combined with the above control schemes for a global exoskeleton robot control scheme. Five subjects performed a peg in hole task for three different target locations to verify the performance of the proposed control scheme. The velocities and interaction forces at the upper arm, lower arm, handle and tip were recorded during the experiments. Power exchange between the subject and device was calculated for performance evaluation. Result shows that proposed control scheme outperforms purely reactive task space admittance control with energy exchange lowered by 11.22 %.

Based on the proposed exoskeleton control scheme, the exoskeleton robot is applied to the stroke patient rehabilitation research project as a clinical trial. In order for this, 3-D video games directly interacting with the robotic system were designed and the assistive force mechanism for the patients was implemented in the exoskeleton robot. In addition a new metric called instantaneous efficiency(IE) was established to evaluate the therapeutic improvement. This metric is designed to reveal the degree of patient's improvement in terms of natural human arm movements by looking at the joint angle configuration, speed and frequency of using uncomfortable joints comprehensively. As their movements are getting closer to the natural human arm movement of normal people, the IE index increases. Ten subjects participated in this pilot research project either as an unilateral or a bilateral therapy group for six weeks. Results shows that even in a short six weeks rehabilitation program, patients' fugl-meyer scores as well as the efficiency index of the movements were significantly improved.