Solver Gurobi Settings and Their Effect on TSP Computational Performance

The Traveling Salesman Problem (TSP) is a classic combinatorial optimization problem widely used as a benchmark for heuristic and exact methods. It determines the lowest-cost Hamiltonian cycle in a complete graph. This work analyzes the impact of Gurobi solver configurations on TSP resolution through a univariate analysis, considering parameters such as presolve, slice generation, heuristics, and MIPFocus. The experiments were conducted in small and medium-sized TSPLIB instances, varying one parameter while the others remain at their default values. Performance was evaluated by computational time. The results indicate specific gains, with more significant improvements in small instances when disabling cuts and greater relevance of mipfocus_2 in medium instances. In general, the effect of parameterization is non-linear and dependent on the characteristics of the instances, reinforcing the importance of adaptive approaches based on meta-learning.