Control of Human Arm Movement: Models and Mechanical Constraints. (Technical Report).

摘要

The first part of this thesis investigates the role of structured models in autonomous motor learning. Any autonomous system, such as the human motor system, has only the internal consistency of its various sensors to rely upon for model building (learning). To study the possibility of learning structured models from internal consistency constraints, the specific problem of learning the kinematic parameters (relative link orientations and length) of general revolute joint manipulators is explored. First it is note that a manipulator may form a mobile closed kinematic chain when interacting with the environment, if it is redundant with respect to the task degrees of freedom (DOFs) at the endpoint. Then it is demonstrated that if the mobile closed chain assumes a number of configurations, then loop consistency equations permit joint angle readings; endpoint sensing is not required. (JS)