Origami folding is one of the most difficult tasks for robots because paper is deformable, thin, and its stiffness changes during folding. In this study, we propose an autonomous vision-based robotic origami system, including the hardware, two folding primitives, and the software.

For the hardware part, a dual-fingered gripper, which has one pneumatically actuated soft finger and one servo-controlled rigid finger, and a spring-loaded slider, which can convert force control to precise position control, are introduced. Our hardware platform is composed of the two end-effectors and two industrial arms-UR10.

Two folding primitives (flex-flip and scooping) are investigated to achieve dexterous paper folding. The flex-flip primitive deals with the deformable linear/planar objects through dyna...[ Read more ]

Origami folding is one of the most difficult tasks for robots because paper is deformable, thin, and its stiffness changes during folding. In this study, we propose an autonomous vision-based robotic origami system, including the hardware, two folding primitives, and the software.

For the hardware part, a dual-fingered gripper, which has one pneumatically actuated soft finger and one servo-controlled rigid finger, and a spring-loaded slider, which can convert force control to precise position control, are introduced. Our hardware platform is composed of the two end-effectors and two industrial arms-UR10.

Two folding primitives (flex-flip and scooping) are investigated to achieve dexterous paper folding. The flex-flip primitive deals with the deformable linear/planar objects through dynamic internal energy exchange, while the scooping primitive lends itself to relatively rigid and thin objects. The two primitives are applied in robotic origami as the stiffness of paper changes during folding.

For the software architecture, several vision techniques are presented and merged. It begins with objectLocalization to track the paper in the real world; after the execution of one folding primitive, it then conducts corner matching with opticalFlowTracing; the system terminates when all the identified creases are folded.

Based on the proposed system, a couple of flat origami patterns, which require sequential straight-line folds, could be achieved with high accuracy. Set the target point as (0,0) mm, the accuracy of the folded point is (-0.750, -0.070) mm for a single flex-flip-based fold, while the accuracy is (0.775, 1.230) mm for a single scooping-based fold. Finally, an airplane origami task and a continuous yacht origami task are performed to show the effectiveness of our proposed system for origami folding.