In precision robotic assembly a small misalignment between parts to be assembled woudl always occur at the interface between mating parts due to positioning inaccuracy of robot, assembly fixtures and manufacturing tolerance of parts. This misalignment, however small can produce large contact forces, resulting in damage to parts or robots and thus lead to failure in assembly. Therefore the misalignment has to be detected and compensated accurately during mating period. For this purpose visual sensing technique has been widely used since it can detect rather large misalignment and also parts shape distance. In case of adopting camera, this technique usually obtain local information due to the limited range of its field of view. Furthermore, it can not avoid self-occlusion generated by the invisible region occluded by objects to be viewed. This problem has been a major hindrance to the success of assembly action when assembly is execute by a vision-based technique. In this paper, two novel visual sensing methodologies are developed to avoid such criticisms. Both system consists of four components: an inside mirror and outside mirror, a pair of plane mirrors and a camera with a collecting lens. The difference between the two is that the system A adopts conic mirror configuration, while the system B employs pyramidal one. Due to this configuration difference, the system A is shown to be capable of detecting two pi omni-directional image, while in addition to this functionality the system B can detect 3D measurement of objects with only one image capture. The measurement principles are described in detail and compared with each other. The image acquiring process is shown to easily detect the in-situ status of assembly action, while the recognition method is found to be effective to identify instantaneous misalignment between the peg and hole. The results obtained from a series of experiments show that the proposed visual sensing methods are an effective means of detecting misalignment between mating parts even in the presence of self-occlusion. The implication is that they will dramatically increase the rate of success when actually utilized in assembly process.
|