Accurate recognition of object contour based on flexible piezoelectric and piezoresistive dual mode strain sensors

The application of flexible wearable sensors in the grasping process of robot hand can recognize the contour, soft and hard, material, surface temperature and other information of the grasping object, which can effectively improve the intelligent level of the robot. In this work, a method of object contour recognition is proposed by combining the flexible PVDF polymer piezoelectric sensor and high conductivity hydrogel piezoresistive sensor aiming at the problem of profile recognition for objects of the same or similar material. The response of flexible piezoresistive sensor to the static strain is used to sense the angular displacement of robot fingers, and then the shape and size of the object is recognized indirectly. At the same time, the flexible piezoelectric sensor is used as the fingertip tactile sensor to reflect the surface morphology of the object through the dynamic strain information when touching the object. In the whole process of grasping the object, the dual-mode strain information is fully used to realize the recognition of the shape, size and surface morphology of the object. Combining these information, the accurate recognition of the object contour can be further realized. In the experiments, six objects with different shape and four objects with different surface morphology are recognized to verify the feasibility of piezoresistive sensors and piezoelectric sensors respectively. In a comprehensive experiment, eight objects made of the same rubber material with different shape, size and surface morphology are recognized, and the average recognition rate is about 84%, which shows good classification advantages for the objects with similar shape, size and material. (c) 2021 Elsevier B.V. All rights reserved.

Automated summary

The research proposes a method for recognizing object contours using flexible sensors in robot hand grasping, enabling recognition of shapes, sizes, and surface morphology by utilizing dual-mode strain information, achieving accurate object contour recognition.