A humanoid manipulator control system and its control method based on deep learning

Abstract

The invention discloses a humanoid manipulator control system based on in-depth learning and a control method thereof. The system comprises an image acquisition module, an object detection and recognition module and a humanoid manipulator control and decision-making module, wherein the image acquisition module is used for achieving an image acquisition and image pretreatment function for a workingscene; the object detection and recognition module is used for detecting target objects in acquired images and recognizing corresponding categories; and the humanoid manipulator control and decision-making module is used for planning a humanoid manipulator motion locus, performing motion decision-making on positions where the target objects are located according to treatment results of the two above modules, and achieving motion locus self-learning. According to the humanoid manipulator control system, an automatic control system is established, self-actuated working can be achieved under conditions without human intervention, the intelligence of machines is improved, the system can be widely applied to environments including factories, workshops, logistics and the like where object sorting and other operation are needed, and the market prospects and the potential value are good.

Description

technical field [0001] The invention relates to the technical field of deep learning image processing and manipulator control, in particular to a humanoid manipulator control system and control method based on deep learning. Background technique [0002] Since the United States Unimation developed the world's first industrial robot in the late 1950s, the application of industrial robots has gradually spread to all aspects of life and production, including but not limited to industrial production, assembly, ocean exploration and development, space exploration, medical treatment Applications and other fields have greatly improved people's living and working conditions. With the continuous development of robot technology, robots have gradually become intelligent. There are more and more service robots for hotel services and home services. The corresponding human-computer interaction technology and robot intelligent control technology are in urgent need of research and developme...

AI Technical Summary

Problems solved by technology

[0005] Based on the above research status, it is necessary to overcome the limitations of traditional perception, interaction and learning methods of industrial robots, such as task determination, offline learning, poor scalability, and inability to adapt to changing environments, and establish a target selection perception method based on selective attention. Influenced human-computer interaction methods and adaptive interactive behavior learning methods enable industrial robots to actively and selectively perceive and process relevant information from production workers, production environments and production objects in dynamic scenes, and online and autonomously learn interactive behavior and previous experience, thereby improving the operational skills of industrial robots and the ability to adapt to unknown or dynamic environments

Images