Online self-learning method for a robot autonomous object pickup task

Abstract

The invention provides an online self-learning method for a robot autonomous object pickup task. The robot stereo camera obtains RGBD information of the stacked to-be-grabbed objects; partitioning theRGB image and the depth map according to a certain step pitch; and inputting the image blocks into an area prediction neural network module in a pairing manner to obtain an absorbability probabilityarea graph, selecting a capture point in an image coordinate system according to the probability graph, converting the capture point to a robot operation coordinate system, executing the capture pointby the robot, and finally training an area prediction network on line according to a robot execution result. According to the method, a deep neural network is utilized to directly obtain grabbing points; According to the method, the problems of object segmentation and pose estimation generally faced by object pickup under a complex background are avoided, meanwhile, the prediction network is trained online by utilizing an execution result of the robot, online self-learning of the object pickup robot is effectively realized, and the problems of complex feature selection or manual marking sample and neural network training are avoided.

Description

technical field [0001] The invention relates to the field of self-learning of robots, in particular to an online self-learning method for autonomous object picking tasks of robots. In particular, it relates to an online self-learning method and system for robot autonomous object grasping involving scattered and piled objects. Background technique [0002] The robot's autonomous learning ability weakens the detailed setting requirements for specific tasks in robot applications and improves the robot's adaptability to different scenarios. It is also an important feature of robot intelligence and has a wide range of application values. There is a wide range of application requirements for the efficient picking of messy and stacked objects in unstructured scenarios, such as automatic sorting of garbage in garbage disposal scenarios, automatic sorting of logistics packages, and automatic loading and unloading in industrial application scenarios. Traditional object picking method...

AI Technical Summary

Problems solved by technology

Traditional object picking methods mostly use a series of processes such as object recognition, object segmentation, and pose estimation to determine the grabbing point to achieve object grabbing. However, stacked objects have complex environmental backgrounds and mutual positional relationships. challenge

At the same time, the variability of application scenarios also brings a huge workload to the specific detailed settings for each scenario.

Images