An intelligent system in which robots perform puzzle matching type assembly job is developed. The robot system is capable of visual recognition and combination for randomly placed jigsaw puzzle parts via coordinated motoins of three robotic arms and conveyor belt, and by employing computer techniques for manipulation of arbitrary geometric patterns.
To be specific, shape representation suitable for matching parts of two-dimensional objects is obtained. In order to characterize pieces, there are proposed a newly defined c-measure to reduce quantization error and a variable smoothing factor to choose critical point correctly. By giving weight according to the sharpness, pieces (slaves) that are "most likely" to mate with given pieces (master) are selected and ordered. Likelyhood of fit is determined and to overcome illegal matching, some heuristic methods such as local/global overlap checking methods are proposed. To check the correctness of the matched figure, the minimum area rectangle encasing an arbitrary closed curve is used as a simplicity measure, and further a method is proposed by which the correctness of the matched figure is determined by the proposed simplicity measure.
The proposed jigsaw puzzle matching algorithm is first proven by computer simulation. To show validity of the proposed algorithm for the applicability to assembly tasks, an intelligent robot system comprising three articulated robots, conveyor belt and vision system has been developed and tested in real time.