Realization of human task skill in dynamic telemanipulation

摘要

Investigates two basic problems in robotics research. It first analyses how a human learns to adapt to his/her unknown environmental constraints during performance of a crank rotation task. It is found from experiments that, as the training proceeded, the subject changes his/her turning strategy from pulling the crank to pushing it. It is clarified that, although the normal contact force between the arm and the crank seems not useful in rotating the crank, it can be used to adjust the arm-crank system's structured stability. When the subject does not familiar with the constraint of the crank, he/she pulls it to make the arm-crank system stable. However, after sufficient training, the subject actively pushes the crank to make the closed chain unstable, which improves the task performance. The paper also proposes a master-slave control approach for the robot system to realize the human's task skill on the telemanipulated object. The approach uses the object's model to compensate the master robot so that the operator can feel as if he/she is working directly with the object. The effectiveness of this approach is also studied experimentally by comparison with two other methods.