Dexterous hand grabbing planning method based on four-stage convolutional neural networks

Abstract

The invention discloses a dexterous hand grabbing planning method based on a four-stage series convolutional neural network, and the method comprises the steps: obtaining a grabbing frame data set anda grabbing gesture data set, carrying out the training of a previous three-stage convolutional neural network and a fourth-stage convolutional neural network, determining the parameters of the network, and obtaining a grabbing model of a dexterous hand, wherein in the proposed four-stage series convolutional neural network, the first three convolutional neural network are used for obtaining an optimal capture box of a target object, and the fourth convolutional neural network is used for predicting the grabbing gesture of the dexterous hand and obtaining various grabbing characteristics through a multi-input network, so that the grabbing gesture in the current state is predicted according to the image information of the grabbed part of the target object and the pose information of the dexterous hand. The unknown object can be grabbed finely, so that the dexterous hand grabs the unknown object without being limited by the unknown object, and the grabbing success rate of the dexterous hand is increased.

Description

technical field [0001] The invention belongs to the technical field of computer vision, and in particular relates to a grasping planning method for a dexterous hand based on a four-level convolutional neural network. Background technique [0002] As the basic function of robots, object grasping operation has always been an important research direction in the field of robotics. In general, grasp planning algorithms are divided into analytical and empirical methods. In order to realize the dexterous hand grasping the object, the analysis method is to determine the grasping pose and gesture of the dexterous hand through the 3D model of the object, but it is usually difficult to obtain the 3D information of the object in the real environment, which greatly limits their application in the actual scene. Applications. Empirical methods based on deep learning have been widely used in grasping planning, but most of these works only consider grasping planning for simple grippers. S...

AI Technical Summary

Problems solved by technology

In order to realize the dexterous hand grasping the object, the analysis method is to determine the grasping pose and gesture of the dexterous hand through the 3D model of the object, but it is usually difficult to obtain the 3D information of the object in the real environment, which greatly limits their application in the actual scene. Applications

Afterwards, empirical methods based on deep learning have been widely used in grasping planning, but most of these works only consider the grasping planning of simple grippers.

Since the dexterous hand cannot be closed directly like a gripper, grasp planning needs to take into account the grasping gestures of the dexterous hand, so they cannot be used in more complex dexterous hand grasp planning

Images