Adaptive acquisition of virtualized deformable objects with a neural gas network

The paper presents a novel approach to guide the acquisition of deformable objects by selecting only a few measurements on the surface of the object. The main idea relies on embedding elastic behavior as a fourth dimension in a neural gas architecture and obtain the sample points as a result of its training. The technique has been successfully applied for objects exhibiting both homogeneous and non-homogeneous elasticity. Early results prove the feasibility and validity of the proposed method.