Controlled Gaussian process dynamical models with application to robotic cloth manipulation

Abstract Over the last years, significant advances have been made in robotic manipulation, but still, handling of non-rigid objects, such as cloth garments, is an open problem. Physical interaction with objects uncertain and complex to model. Thus, extracting useful information from sample data can considerably improve modeling performance. However, training models a challenging task due high-dimensionality state representation. In this paper, we propose Controlled Gaussian Process Dynamical Models (CGPDMs) for learning high-dimensional, nonlinear dynamics by embedding them low-dimensional manifold. A CGPDM constituted latent space, associated where external control variables act mapping observation space. The parameters both maps are marginalized out considering priors. Hence, projects high-dimensional space into smaller dimension which it feasible learn system data. capacity has tested simulated real scenario, proved be capable generalizing over wide range movements confidently predicting motions obtained previously unseen sequences actions.