Method for quickly processing "kidnapping" problem of mobile robot positioning

Abstract

The invention discloses a method for quickly processing a "kidnapping" problem of mobile robot positioning, and belongs to the technical field of mobile robots. The method comprises the following steps of: training a convolution neural network model based on deep learning through training data collected in advance to obtain a rough pose of a camera at the time of taking a picture by inputting thepicture only; when the robot judges that there is a "kidnapping" problem in itself, sending a picture taken by a camera on the robot at the moment to the convolution neural network model to obtain a rough pose; and then adopting a particle filter positioning algorithm in a specific format to reduce the number of iterative convergence in the algorithm and achieve a goal of quickly solving the "kidnapping" problem. According to the method for quickly processing the "kidnapping" problem of the mobile robot positioning, the mobile robot can be quickly and accurately repositioned after the robot ismanually moved to any position.

Description

technical field [0001] The invention relates to the field of robots, and in particular to a method for rapidly handling the problem of positioning "kidnapping" of a mobile robot. Background technique [0002] The localization of mobile robots is an important research hotspot in the field of robotics. The "kidnapping" problem will be encountered in the positioning process of the mobile robot, that is, the mobile robot passively moves to any position on the map, which can be simply understood as artificially moving the robot to any position, but the robot thinks that it is still at the origin and the positioning fails The problem. Since the GPS signal has no signal in some areas or indoors, how to relocate when the "kidnapping" problem occurs is an urgent problem to be solved, and most current positioning algorithms cannot quickly solve the robot "kidnapping" problem. [0003] With the development of deep learning technology and computer vision technology, many methods for p...

AI Technical Summary

Problems solved by technology

The literature "Research Progress in Localization and Map Creation of Mobile Robots Based on Particle Filters" (Robotics, 2007, 29(3): 281-289, 297. Yu Hongshan, Wang Yaonan) introduces some methods of robot map creation and localization using particle filters , but in a complex environment due to the similarity of the scene structure, a single particle filter localization is likely to cause localization failure without any prior knowledge of the robot position

[0004] There are certain shortcomings in the above methods: 1) The accuracy of the convolutional neural network model based on deep learning is about 2m and 5°, which cannot achieve accurate positioning

2) Although the particle filter positioning algorithm can solve the "kidnapping" problem, it needs to go through multiple resampling, which takes a long time, resulting in a decrease in real-time performance

Images