Solver CPLEX, version 12.6, was used to run the mathematical model on a computer with an
Intel Core i7 processor with 16GB of RAM. The maximum execution time was limited to
7,200 seconds for each instance. When running instances with CPLEX it was possible to obtain results to measure the quality of the RAS metaheuristic solution. RAS was developed in C language and run on the same computer described before with a runtime limited to 7,200 seconds. RAS metaheuristic was run 10 times for each analyzed instance to obtain the deviation from the best solution found.
The results achieved by CPLEX and by the proposed RAS metaheuristic for the 14 test instances and for the 6 Real Case instances are shown in Table 2. The Instance column shows the instances analyzed by the CPLEX and the RAS. The five subsequent columns show the results achieved by CPLEX. The FO column shows the value of the Objective Function (FO), corresponding to the transportation total cost. The UB and LB columns show, respectively, the upper and lower bound found when CPLEX was not able to reach the optimal solution of the instance. The Exec. …show more content…
Another interesting observation is that after running these same instances in CPLEX without applying the lower and the upper bounds based on the number of buses (proposed in this paper in Section 4), CPLEX stopped after 7,200 seconds with a GAP of 99.94% and 99.95%, respectively. With the introduction of these bound, for the same instances, CPLEX stopped also after 7,200 seconds with a GAP of 0.05% and 0.03%, respectively, showing the importance of the proposed bounds for the mathematical