Owing to the flexibility and redundancy of neuromuscular and skeletal systems, humans can trace the same hand trajectory in space with various arm configurations. However, the joint trajectories of typical unrestrained movements tend to be consistent both within and across subjects. In this paper we propose a method to solve the 3-D inverse kinematics problem based on minimizing the magnitude of total work done by joint torques. We examined the fit of the joint-space trajectories against those observed from human performance in a variety of movement paths in 3-D workspace. The results showed that the joint-space trajectories produced by the method are in good agreement with the subjects' arm movements (r2>0.98), with the exception of shoulder adduction/abduction (where, in the worst case, r2 approximately 0.8). Comparison of humeral rotation predicted by our algorithm with other models showed that the correlation coefficient r2) between actual data and our predictions is extremely high (mostly >0.98, 11 out of 15 cases, with a few exceptions, 4 of 15, in the range of 0.8-0.9) and the slope of linear regression is much closer to one (<0.05 distortion in 12 out of 15 cases, with only one case >0.15). However, the discrepancy in shoulder adduction/abduction indicated that when only the hand path is known, additional constraint(s) may be required to generate a complete match with human performance.

No abstract

A computational model of a learning system (LS) is described that acquires knowledge and skill necessary for optimal control of a multisegmental limb dynamics (controlled object or CO), starting from "knowing" only the dimensionality of the object's state space. It is based on an optimal control problem setup different from that of reinforcement learning. The LS solves the optimal control problem online while practicing the manipulation of CO. The system's functional architecture comprises several adaptive components, each of which incorporates a number of mapping functions approximated based on artificial neural nets. Besides the internal model of the CO's dynamics and adaptive controller that computes the control law, the LS includes a new type of internal model, the minimal cost (IM(mc)) of moving the controlled object between a pair of states. That internal model appears critical for the LS's capacity to develop an optimal movement trajectory. The IM(mc) interacts with the adaptive controller in a cooperative manner. The controller provides an initial approximation of an optimal control action, which is further optimized in real time based on the IM(mc). The IM(mc) in turn provides information for updating the controller. The LS's performance was tested on the task of center-out reaching to eight randomly selected targets with a 2DOF limb model. The LS reached an optimal level of performance in a few tens of trials. It also quickly adapted to movement perturbations produced by two different types of external force field. The results suggest that the proposed design of a self-optimized control system can serve as a basis for the modeling of motor learning that includes the formation and adaptive modification of the plan of a goal-directed movement.

We recorded neural activities in dorsal premotor areas (PMd) when monkey was preparing to perform arm reaching movements in motor-set period before his actual execution. He was required to select one of four possible movements (reaching a target on the top-left or top-right corner of a virtual cube, moving straightly or making a detour) in accordance with two sets of instruction cues. Among the neurons recorded, we analyzed 118 neurons which showed set activities during movement preparation. More specifically, 46 neurons were modulated with respect to the obstacle-related conditions;42 neurons, with respect to target location and 15 were tuning to both. In trials in which the virtual obstacle appeared, successful trials were also compared to unsuccessful (hit-obstacle) trials, in which the monkey moved straight to the target instead of bypassing the virtual obstacle. The motor-set activity of a specific group of neurons was significantly different in set activities between successful and unsuccessful trials. Their activity was strongly modulated by the spatial position of the virtual obstacle, indicating that those cells likely participated in the planning of the hand path (straight or curved in counterclockwise direction) based on visual information about the virtual obstacle. At the same time, neurons in which the spike activity modulation with respect to the virtual obstacle position was relatively weak, did not show significant differentiation of the motor-set, preparatory activity between successful and unsuccessful trial types.

To generate a naturally looking arm movement when using cortically derived hand trajectory is investigated. One approach to solving this task is through the utilization of the elbow swivel angle, based on the assumption that its value changes insignificantly with relation to different target locations. The goal of this study is to test experimentally whether the swivel angle significantly depends on the horizontal and vertical coordinates of the target in case of an unrestraint voluntary reaching movements in 3D. In out experiment, the subjects were required to move their hand to different targets located on a vertical board in front of the subject. The results show that the swivel angle depends systematically on both vertical and horizontal coordinates of the target location. The average range of swivel angle variation is approximately 25 degrees. Thus, the assumption that the swivel angle is approximately the same for different target locations is inadequate. Different methods for determining the exact angular configuration of the arm (e.g. those based on the optimization approach) should be employed.