A Genetic Algorithm Combined with Stacked Denoising Sparse Autoencoders

Abstract

The invention designs a genetic algorithm combined with a stacked noise-reduction sparse autoencoder, which mainly includes a stacked noise-reduction sparse autoencoder part, a SOM neural network part and a genetic algorithm. The real-time environmental image features are extracted by stacking noise-reduction sparse autoencoders, and the influence of environmental noise is eliminated at the same time. The mapping fitting of the SOM neural network is used as the fitness value in the genetic algorithm, which solves the problem when the traditional genetic algorithm is applied to complex real-world environments. The lack of flexibility and accuracy of the fixed fitness calculation function avoids the problem of deception in the algorithm and improves the quality of the algorithm solution. At the same time, the SOM neural network mapping fitting can effectively avoid the long-term iterative operation and output error problems of other neural networks.

Description

technical field [0001] The invention relates to the field of mobile robots, in particular to a genetic algorithm combined with a stacked noise reduction sparse autoencoder. Background technique [0002] With the development of robot technology, robots have begun to be applied to unknown environments. Compared with the research on task assignment of mobile robots in known environments, robots in unknown environments do not have prior knowledge of the environment and need to process perception information from the real environment. , Uncertain and incomplete information of the environment, it is inevitable to encounter various situations in the process of multi-task assignment. Therefore, how to improve the adaptability of mobile robots to the environment and improve the traditional idealized task allocation algorithm has very important practical significance. [0003] The traditional genetic algorithm is applied to the multi-task assignment of robots. Due to improper selecti...

AI Technical Summary

Problems solved by technology

[0003] The traditional genetic algorithm is applied to the multi-task assignment of robots. Due to the improper selection of the fitness function or the fixed function does not conform to the actual application scenario, the following deception problems are prone to occur: 1. In the early stage of the genetic algorithm, some supernormal individuals are usually produced, and these supernormal individuals will The selection process is controlled due to the outstanding competitiveness, which affects the global optimization performance of the algorithm; 2. In the later stage of the genetic algorithm, when the algorithm tends to converge, due to the small difference in the fitness of individuals in the population, the potential for further optimization is reduced, and a certain local optimum

Images