Automatización de la asignación de espacios dentro de una bodega semiautomática para la empresa Robotrón a través de inteligencia artificial

Abstract

Given the growing and demanding technological advancement, applications and techniques of artificial intelligence in the industry, the automation of logistics processes, added to the scarce implementation of intelligent systems in warehouses of small and medium-sized companies, the opportunity arises to design, test and implement an algorithm based on artificial intelligence (Deep learning) using cutting-edge technology and tools. This project is focused on the development of software using the LSTM neural network architecture, based on previous studies and free software, in order to make predictions based on historical product data, which allow for area segmentation. and physical space reserves, in turn of an allocation algorithm which designates a single space within the previously reserved areas. By implementing the software on a server of the Robotrón company, making use of data from 10 clients of an external company and 30 sales products, it was possible to validate the effectiveness of the architecture designed for the input data with a maximum RMSE error of 3.49. Within the experimental validation tests, a minimum error of 1.13% was obtained in the first month of prediction and a maximum error of 12.3% in the second month of prediction. In turn, by making use of the intelligent software, it was possible to reduce the time of allocation and entry to the warehouse by up to approximately 6 minutes.