Robot skill acquisition method based on meta-learning under guidance of scene memory

Abstract

The invention provides a robot skill acquisition method based on meta-learning under the guidance of scene memory. The method comprises the following steps: firstly, establishing a robot learning system scene memory model, and constructing a robot perception and memory similarity measurement algorithm to realize event and scene information retrieval matching and event update calling in scene memory; and constructing a robot operation skill meta-learning algorithm guided by scene memory, obtaining knowledge from the independent tasks and all the tasks, and performing skill learning. According to the method for guiding the robot to learn new skills by using the existing experience, the learning efficiency of the robot on the operation skills is improved, and the problems that in the robot operation skill learning process, the data size is too large, and similar tasks need to be trained repeatedly are solved.

Description

technical field [0001] The invention belongs to the technical field of intelligent robot services, and relates to a robot operation skill learning method based on scene memory and meta-learning. Background technique [0002] In recent years, intelligent robots have been used in the fields of industrial production, medical care, commerce, and home services. The current robot learning methods are capable of performing precise and repetitive tasks, but they lack the ability to learn new tasks. Similar task scenarios require repeated training. Inability to accumulate experience to guide new tasks to achieve rapid learning and other issues. In the invention patent CN108333941A, Du Guanglong and Zhang Ailing of South China University of Technology disclosed a cloud robot collaborative learning method based on hybrid enhanced intelligence. It uses the meta-learning method of neural task programming to decompose the total task into simple sub-tasks, and the robot learns the sub-tas...

AI Technical Summary

Problems solved by technology

[0003] The main problem to be solved by the present invention is how to use learned knowledge and existing experience to solve new tasks faced by intelligent robots and adapt to new task goals.

Aiming at the current robot skill learning that requires a large amount of data training, similar task scenarios need repeated training, and cannot accumulate experience to guide new tasks to achieve rapid learning, etc., the present invention proposes a meta-learning robot skill learning method combined with episodic memory

Images