Results and discussion
Sensitivity and what-if analysis are generally required in complex decision problems. In vehicle routing problems it is usually interesting to obtain answers about the impact of changes on the required resources (e.g., duration of the crews' work shift, capacities of vehicles and respective costs), or on the network (e.g., modifying traversal times or temporarily deleting arcs that become not available for truck access during road maintenance or other causes). The urban trash collection system planner may want to evaluate if changes in vehicle capacity and/or shift times affect the number of vehicles and routes required. The analysis of such trade-offs are important in the management of vehicle fleets and required manpower. In our example, results were obtained (solution S1) by solving the problem with a shift limit of 6 h (Table 1) for two types of vehicles with different capacities. Five routes were generated, organized in three shifts for vehicle type 2 and one shift for vehicle type 1. The objective was the minimization of the total traveled vehicle length.
Solving again the problem by minimizing the number of trucks instead the total traveled length (with the remaining parameters unchanged), a similar solution to S1 (in terms of total length traveled, number of required shifts, and vehicles) was obtained (solution S2) with five routes, organized in four shifts (one for vehicle type 1, and three for vehicle type 2). However, increasing the shift limit from 6 to 7 h, a different solution (S3) was obtained, with five routes organized in three shifts (one for vehicle type 1, and two for vehicle type 2), with a slightly higher total length traveled (+ 1,2 %); see Table 2. In all the three solutions mentioned above equal costs were assumed for both vehicle types. Maintaining the shift limit of 6 h, minimizing the total traveled vehicle length and considering vehicle type 2 with a relative cost 50% higher than the cost of vehicle type 1, the system generates a solution with five shifts using only vehicle type 1. In this case, the total traveled vehicle length obtained was 9% higher than in S1.