A new neural architecture based on ART and AVITE models for anticipatory sensory-motor coordination in robotics

摘要

In this paper a novel sensory-motor neural controller applied to roboticsystems for reaching and tracking targets is proposed. It is based on howthe human system projects the sensorial stimulus over the motor joints, sendingmotor commands to each articulation and avoiding, in most phases of the movement,the feedback of the visual information. In this way, the proposed neuralarchitecture autonomously generates a learning cells structure based on theadaptive resonance theory, together with a neural mapping of the sensory-motorcoordinate systems in each cell of the arm workspace. It permits a fast openloopcontrol based on propioceptive information of a robot and a precisegrasping position in each cell by mapping 3D spatial positions over redundantjoints. The proposed architecture has been trained, implemented and tested in avisuo-motor robotic platform. Robustness, precision and velocity characteristicshave been validated.