Optimization of the inventory routing problem with Artificial Intelligence

Dissertação de mestrado em Informatics Engineering ; Nowadays the success of a business is dependent on the ability to effectively integrate in an intricate network of entities that are connected by material and information flows, inventory management being one of the main concerns. These flows are characterized by decision-making processes that will vary depending on the environment, entities and business models in the network. So, these networks need a decision-making system capable of providing solutions that dictate the optimal way the network and its entities provide and collect inventory in order to reduce costs and maximize profit. In the context of this dissertation, the problem arises when there is a stock disruption in the network and outside entities can no longer answer the stores’ supply requests and these stores become the entities responsible for requesting and delivering products to each other. This problem is modeled as an Inventory Routing problem, since it encompasses inventory management and routing decisions. The main goal of the system can be described as maximizing the collection of products per travel distance, without causing stock-outs at any supplier, for the entire network. The problem at hand is an optimization problem. In order to solve this optimization problem, first, the structural characteristics and key aspects were identified and studied, followed by the mathematical conceptualization, which involved the definition of the objective function and the corresponding set of constraints. The mathematical formulation allows the problem to be translated into a specific and precise mathematical language, making it possible to evaluate solutions, by means of a fitness function, and apply optimization algorithms to solve the problem. These optimization algorithms can be approximate or exact methods and their suitability to the problem depends on many factors such as the size, structure and complexity of the problem. So, the choice of the optimization algorithms must be preceded by a ...