Inteligência artificial e computação quântica para solução de problemas de logística industrial

Abstract

The warehouse management problem discussed in this dissertation consists of optimizing the allocation of items in a storage system ordered similarly to a queue and minimizing reallocations of these items. For this purpose, this research proposes three strategies: Quantum Annealing, Simulated Annealing and a Recommendation System, the latter also responsible for generating parameters for the other methods. The results showed that the Recommendation System stood out in terms of processing time and significantly reduced reinsertions. On the other hand, Simulated Annealing achieved even better results, with a significant reduction in the number of reinsertions compared to the company's current method. Both methods revealed practical potential and viable integration into industrial systems. In the quantum approach, despite restrictions on the instances executed, the evolution of the system's energy indicates competitive potential in relation to Simulated Annealing, especially considering the expected advances in fault-tolerant computers. Although currently quantum computers are limited and susceptible to a considerable level of noise, the results obtained already point to practical and promising applications. In addition, a connection was identified between the Lift metric for association rule and the g^(2)(τ) correlation function of quantum optics. On this basis, a study was started to verify the feasibility of developing a quantum recommendation algorithm, whose initial discussions are presented at the end of the dissertation.