A multiple internal model approach to movement planning

Under the hypothesis that any human motion could be decomposed into dynamic movement primitives (DMPs), sets of second order differential equations are used as the internal model (IM) to describe primitive movements. The spacial and temporal scalabilities of the internal model could be used to simplify the learning process. In this paper, we present an approach to movement learning based on internal models. By making use of the linear properties of internal models, we first investigate the possibility of generating similar movement patterns directly via the same internal model with the minimum changes in the internal model parameters, and avoid the reinforcement learning. Next, we consider more complex movements for which different internal models are needed. Based on the task decomposition, all movements can be classified into the sequential and parallel DMPs. The former requires a number of IMs to work sequentially so that a complicated movement can be performed. The latter also requires a number IMs to work in parallel to generate the needed movement. To mimic the human limb behavior, we use a two-link robot arm as the prototype to perform the movement in the process of letter writing.