Direct Perception for Intelligent Robot to Grasp an Unknown Object

• 作 者: HiroyukuMasuta;Hun-okLim;
• 机 构: Kanagawa University,Japan
• 出 处: 第三届先进计算智能与智能信息学国际研讨会(The Third International Workshop on Adv
• 发表日期: 2013年期

　　This paper describes a robot perception of an unknown object based on perceiving-acting cycle concept for service robots using 3D range sensor.The previous method has problems which are high computational costs and low-accuracy of small object detection to detect an unknown object.We have proposed retinal structure based plane detection method to solve aforementioned problems.The proposed method is constructed by retinal structure based simple plane detection and integrated object plane detection.Furthermore,we have proposed the sensation of grasping based on ecological psychology to perceive the possibility of grasping an object,even if a robot doesnt know the accurate properties of an object.The sensation of grasping affords the possibility of action to a robot directly without inference from physical information such as size,posture,and diameter.As experimental results,we show that the computational cost of the proposed method is less than compared to traditional plane detection method.And,we show that the robot can detect a fallen cup that is unknown object,and then grasp a fallen cup.