When DMP Breaks: A Contribution on the Tenacity of Dynamic Movement Primitives from the Handwriting Perspective

Dynamic Movement Primitives are an encoding method oriented to model the generalized behavior of a variable or set of variables over time, providing reproducibility and versatility. Originally designed for motor behavior, the method has demonstrated to be a highly promising tool for modeling tasks for robotic manipulators. An interesting use in literature is handwriting encoding. Close-shape graphemes are nearly impossible to model with existing DMP formulations. In this paper, we explore the operational definition of the method’s classic formulation in consideration, as well as a more tenacious form, focusing on cases which might not be possible to model. We then propose a solution to avoid such cases. Our modified formulation yields more consistent and coherent results regardless of the nature of the modelled handwritten graphemes.