Quick imitation learning method, system and equipment for robot skill learning

Abstract

The invention belongs to the technical field of robots and artificial intelligence, particularly relates to a quick imitation learning method, system and equipment for robot skill learning, and aims to solve the problems that teaching data collection of an existing robot is time-consuming and complex, and when a demonstrator provides suboptimal teaching, the robot cannot optimize the execution process, and the teaching efficiency is high. The operation efficiency of the robot is reduced, and the operation time is prolonged. The system method comprises the following steps: collecting original teaching data; in combination with a pre-trained evaluator, extracting key frame teaching data from the original teaching data set through a preset key frame extraction method; and based on the key frame teaching data, controlling the robot to reproduce teaching through the trained control strategy model. According to the method, the data collection process during robot imitation learning can be simplified, the teaching track is optimized based on the performance of the robot control strategy model, the data collection time and the robot operation time are shortened, the efficiency of robot imitation learning is effectively improved, and the operation time is shortened.

Description

technical field [0001] The invention belongs to the technical field of robots and artificial intelligence, and particularly relates to a rapid imitation learning method, system and equipment for robot skill learning. Background technique [0002] Imitation learning is a method in which robots acquire skills by imitating the behavior of experts. The imitation learning method can be divided into two stages, namely teaching data collection and policy generation. When teaching data collection, the presenter needs to perform one or more tasks and provide status and action information during the presentation. During policy generation, the robot uses a policy generation method to learn a control policy for imitating the presenter's behavior from the taught data. [0003] There are still some differences between the existing imitation learning methods and the human imitation process. It is mainly reflected in: first, the human imitator can only learn skills from the state sequenc...

AI Technical Summary

Problems solved by technology

[0004] In order to solve the above-mentioned problems in the prior art, that is, in order to solve the time-consuming and complicated teaching data acquisition of the existing robot and when the teaching operator provides sub-optimal teaching, the robot cannot optimize the execution process, resulting in a decrease in the operating efficiency of the robot, For the problem of longer operating time, the first aspect of the present invention proposes a fast imitation learning method for robot skill learning, which includes:

Images