Robot route planning method by employing improved convolutional neural network based on K mean value

Abstract

The present invention provides a robot route planning method by employing an improved convolutional neural network based on a K mean value. The method comprises the following three steps: the step 1, establishing a K mean value clustering algorithm model and a BP network; the step 2, inputting the initial patrol point position of the robot into the K mean value clustering algorithm model; and the step 3, taking the output result of the K mean value clustering model as the input value of the BP network, and performing training, wherein the output value of the BP network is namely the optimal robot patrol route. According to the method provided by the invention, the first layer of the network employs K mean value unsupervised learning, and the second layer of the network employs the BP algorithm supervised learning to increase the correlation of the features obtained through the training and the actual tasks, improve the accuracy, reduce the parameters required training and solve the gradient dispersion problem. The improved algorithm is applied to the robot route plan and has high accuracy and faster training operation speed compared to the main algorithms about the experiment results.

Description

technical field [0001] The invention belongs to the technical field of automatic control, in particular to a robot path planning method based on K-means improved convolutional neural network. Background technique [0002] In recent years, deep learning has received more and more attention, and it is regarded as a key link leading to artificial intelligence. As a paradigm of deep learning, convolutional neural networks have been greatly promoted and developed. In 2012, the team led by Krizewski used the deep convolutional neural network to emerge in the world's top event image network large-scale visual recognition challenge, broke the record, and the final score far exceeded the results of other algorithms. In many current recognition tasks, convolutional neural networks have irreplaceable advantages compared with other human algorithms of the same recognition level. Therefore, convolutional neural networks are of epoch-making significance for solving deep learning problem...

AI Technical Summary

Problems solved by technology

[0005] The technical problem to be solved by the present invention is that the existing convolutional neural network has a large number of parameters, the parameters cannot be learned to the optimal value, and the problem of gradient dispersion

Images