A model for solving vehicle scheduling problems: a case of study

In this work we formulate a model to solve a type of vehicle scheduling problem, derived from the operation of the mass transit system (MIO) in Cali city, Colombia. Four companies operate the system with 3 types of buses and four depots. Two kinds of tasks should be assigned to operators’ buses. A task is a sequence of consecutive travels of a route between two stations: initial and final. Each task should start and should finish in a depot, not necessarily the same. There are two main objectives defined by the operators. One objective is to minimize the total deadhead kilometers between depots and stations where tasks should start or end. The other objective is to minimize the maximum deviation of kilometers (commercial and deadhead) assigned to the operators regarding the ideal quantity of kilometers that they should have, according to the number of their buses in the fleet. We have implemented the proposed model in AMPL using Gurobi solver and we validate it using nine representative instances, generated with real data obtained from the system operation. The results obtained, in all cases, improve the solutions proposed by the company. Model variations are proposed to deal with new desirable constraints for the company