Enhancing Movement Smoothness by Using a Human HAnd Trajectory Generation Model With Task-Related Constraints

Natural motion produced by the biological motor control system presents movement smoothness, but neurological disorders or injuries severely deteriorates motor functions. This paper proposes a robot-aided training methodology focusing on smooth transient trajectory generation by the arm while performing a complex task (i.e., virtual curling). The aim of the proposed approach is that a trainee should be taught a reference velocity profile with high movement smoothness in the complex task via the interaction with a robotic device while improving coordination ability for natural arm movements. In the virtual curling training, a trainee manipulates the handle of an impedance-controlled robot to move a virtual stone to the center of a circular target on ice while predicting transient behaviors of the released stone. First, a reference hand motion is clarified through a set of preliminary experiments for different task conditions carried out with four well-trained subjects, and the characteristics of skilled hand velocity profiles are coded with a set of quantitative factors as task-related constraints. The skilled hand motions according to task conditions are successfully simulated in the framework of a minimum-jerk model with the task-related constraints. Next, the training program for enhancing movement smoothness is developed using the computational model, which has four training modes of operation: 1) diagnosis, 2) teaching with active-assistance by the robot, 3) training with passive-assistance, and 4) training with no assistance. Finally, training experiments with ten novice healthy volunteers demonstrate that the proposed approach can be utilized in the recovery of motor functions necessary for desired velocity profiles with high motion smoothness.