Development of a robot hand with four primitive motions for efficient assembly of various ingredients in Bento boxes

In recent years, various robot hands have been developed for handling food, yet arranging diverse foods within a Bento lunch box remains a challenge due to their lack of dexterity and thick fingers. Conventional robot hands, designed for general food handling, fail to perform well in the precise and varied tasks required for Bento preparation. To address this issue, the authors propose a novel robot hand equipped with an additional degree of freedom for linear motion, augmenting the conventional 2-jaw gripper. Our proposed design enables the execution of four primitive motions: ‘grasping,’ ‘releasing,’ ‘pulling,’ and ‘sliding off,’ which the authors propose are essential for the effective arrangement of ingredients in a Bento box. In this study, the authors conducted extensive motion experiments using a single-arm robotic system combined with an image recognition system based on the Segment Anything Model. The results demonstrate that the proposed robot hand can perform the four primitive movements, overcoming limitations of conventional robot hands. Notably, the ‘pulling’ motion facilitated the grasping of thin food items, previously challenging with conventional ‘grasping’ techniques. The proposed robot hand successfully managed all food items within the Bento box used in the experiments, showcasing its potential for practical Bento assembly.