3D Grasp Synthesis Based on a Visual Cortex Model

Abstract

In this paper, the problem of object grasping is considered from both a biological and an engineering point of view. A model of information processing for the grasp synthesis and execution is described based on recent findings from neuroscience. Taking into account the differences between robotic and biological systems, this paper proposes the adaptation of that model to the peculiarities of a robotic system, instead of mimicking it. For this purpose, an architecture is proposed that allows the scalability of this model and its integration within more complex tasks. The grasp synthesis is designed as integrated within the extraction of a 3D object description, so that the object reconstruction is driven by the needs of the grasp synthesis. The integration is formulated as a framework where different grasp synthesis strategies could be applied