Deep-reinforcement-learning-based indoor robot scene active recognition method

Abstract

The invention provides a deep-reinforcement-learning-based indoor robot scene active recognition method, and belongs to field of machine learning and the technical field of robots. The method comprises the steps that a classification neural network NL capable of recognizing sonar information binaryzation contour figure ring projecting vectors is trained; a reinforcement learning training phase isentered, wherein scene recognition testing is conducted on a robot in a scene multiple times, and in the testing process, a reinforcement learning neural network NQ is trained and subjected to fittingto obtain a function approximator; when reinforcement learning neural network NQ finishes training, an execution phase is entered, wherein the robot indoor scene active recognition function is testedaccording to scene contour information collected by a sonar sensor. According to the method, on the basis of an extreme learning machine algorithm, the calculation efficiency is improved; on the basis of a reinforcement learning algorithm, the scene recognition accuracy rate is increased. The method can adapt to different scene recognition tasks, people does not need to be involved, the robot learns actively, and the scene recognition accuracy rate is increased automatically.

Description

technical field [0001] The invention relates to an active recognition method for an indoor robot scene based on deep reinforcement learning, which belongs to the field of machine learning and the field of robot technology. Background technique [0002] In recent years, robots have been more and more used in production and life, such as risk elimination, military detection, medical care, etc.; while the positioning and navigation of robots play a key role in realizing the above functions, quickly and accurately identify robots The environment is the prerequisite for accurate positioning of the robot. In the current scene recognition application, the robot trains the classification neural network through the sample data set collected by the sensor, which can only realize the passive recognition of the current scene; due to the limitation of the robot's orientation and the data limitation of the low-cost sensor, the accuracy of scene recognition is low. low. [0003] In a pri...

AI Technical Summary

Problems solved by technology

The disadvantage of this device is that the device is relatively expensive, the structure is relatively complicated, and the amount of image information collected by the camera is huge, resulting in slow calculation speed.

Images