Intelligent Bionic Manipulator Based on Multimodal Tactile Perception

Bionic manipulator after several generations of development has entered a relatively mature stage, but by the influence of its tactile perception and control system, the universality and complexity of bionic manipulator task still has many problems. The bionic manipulator with only one single mode sensor separates the correlation between each mode and is still lacking in terms of hard grasping. After continuous research and a large number of experiments and tests, a multi-modal fusion hard grasping bionic manipulator is designed in this paper. Firstly, the original data of multiple sensors are filtered, fused and feature extracted. Then K-means clustering algorithm is used for feature level fusion to make a comprehensive decision on the data. So the soft and hard attributes of the captured object can be determined, and then adopt different capture strategies. To further improve the accuracy of grasping objects, Yolo2 neural network visual is used to capture the type, size, shape and position of objects. The experiment shows that the bionic manipulator can grasp the object more accurately, and the damage to the object is less, and the grasp is more stable.