Methodology of an Inverse Kinematic Model for Estimating Shoulder Girdle Angles without Acromial Sensing
The angles of the shoulder girdle are of particular interest when evaluating extravehicular spacesuits for restricted upper limb motion and possible causes of shoulder discomfort and injury. However, in-suit sensor volume limitations, magnetic distortion from suit materials, and substantial skin motion artifacts hinder existing measurement techniques. Several regression models have been developed for estimating these angles from more accessible measurements, but these models have been shown to display limited generalization. A model that can provide better estimates of shoulder girdle angles inside the suit will improve assessment and decision making for hardware and fit evaluation. We have developed an inverse kinematic model of the shoulder girdle for estimating sternoclavicular and scapulothoracic orientations. The model uses skeletal geometry, humerothoracic kinematics, and scapulothoracic separation distances as inputs. Select segment lengths and the location of the glenohumeral joint center are used to determine geometric closure constraints for a mechanically analogous spatial linkage. The resulting inverse kinematics problem is solved with a nonlinear least squares optimization according to the scapulothoracic distances. The performance of the model was tested for a collection of range of motion and task-oriented movements. In ideal simulations, the 50th and 95th percentile RMSE for scapular upward rotation over all movements were 0.3° and 1.5°, respectively. In comparison, the tested regression models returned errors greater than 7.5° and 21°. However, the performance of the linkage model was diminished after simulating small geometry perturbations or measurement noise. For example, reducing the length of the clavicle by 1.8 mm causes the linkage model to produce errors of 8.9° and 15.4°. This is likely due to a lack of compliance in the ideal mechanical joints of the linkage model. Ongoing work includes reducing the model’s sensitivity to these factors by incorporating an additional optimization routine that allows for limited changes in certain segment lengths.