Motor control-learning model for arm's reaching movement without desired trajectory

摘要

In this paper, we propose a motor control-learning model that generates arm's reaching movements without desired trajectories. The model consists of three different neural networks implementing an inverse statics model (ISM), feedback controller (FBC) and forward dynamics model (FDM) of the arm. The ISM transforms an arm's posture into a set of motor commands that shifts the arm's equilibrium to that posture. In our model, final target posture of the movement is inputted to the ISM from beginning to end of the movement. Therefore, the ISM generates constant motor commands throughout the movement in our model and resembles the control hypothesis referred to as "final position control" (Feldman, 1966; Polit and Bizzi, 1979). This control hypothesis does not require desired trajectory for the movement and is considered as the simplest version of "equilibrium-point control hypothesis". Unfortunately, it was rejected by deafferented monkey's experiment suggesting the existence of desired trajectory (Bizzi et al., 1984). In this paper, we show, however, that the monkey's experimental result can be explained without assuming the existence of desired trajectory if the FBC and the FDM are involved in motor control. We also show that our model reproduces human subject's reaching movements in sagittal plane without specifying desired trajectories.