Development of vision guided algorithms for compliantly supported robotic peg in hole assembly

Abstract

The robotic manipulators have been used to automate the time-consuming assembly process and meet out the market demand. The robotic assembly systems lose their flexibility in case of change in assembling components and/or environment, due to their poor sensing and perceiving capabilities. Hence, sensors like vision, depth and force sensors are used to improve the sensing, perceiving, analyzing, and decision making capabilities of a robotic assembly system. A Vision Guided Robotic Assembly System (VGRAS) provides higher flexibility and adaptability to variation in assembly environment, as it does not require parts delivery system to locate, manipulate and align the components. In the fixture-less peg-in-hole assemblies in smart manufacturing industries, the location of the hole component is dynamic and it varies for each assembly. In this concern, a robotic assembly in an unstructured environment is executed in three phases: locating phase, aligning phase and insertion phase. In the location phase, the pose of the hole component is identified using vision sensor, followed by alignment of the peg with the hole components by controlling the manipulator with the aid of vision feedback in the alignment phase. Finally, the peg is inserted within the hole component with the compliance assistance, during the insertion phase. The success of the assembly process is ensured, when the peg component is inserted completely in the hole component without jamming or damaging the mating components and with minimal insertion force and assembly time. Hence, the success of the assembly process is significantly influenced by the pose estimation accuracy in the locating phase, visual servoing stability and its accuracy in the alignment phase, and the appropriate selection of the compliance device for the insertion phase. newline