Sequencing of Containers on Vessels via Multiobjective Optimization with the Use of Entropy and Experimental Validation

The planning of loading of river vessels constitutes a combinatorial problem in which hydrostatic stability and logistic cost represent conflicting objectives. In highly constrained stevedoring problems, multiobjective algorithms can suffer from loss of diversity and premature convergence. In this work, a hybrid approach to the container stowage problem is proposed, integrating the NSGA-II algorithm with a gluttonous repair heuristic and a hypermutation operator guided by Shannon Entropy. Local entropy is used to identify structurally irregular regions of the Pareto Boundary and target adaptive perturbations. In 60 independent executions, the proposed approach showed statistically significant superiority (p < 0.001), dominating approximately 70% of the solutions in the base model and reducing the IGD indicator by about 30%. The experimental validation in a physical prototype confirmed the practical feasibility of the solutions obtained.