André Gustavo dos Santos

Hybrid metaheuristic for the single vehicle routing problem with deliveries and selective pickups

This paper presents a hybrid metaheuristic for the single vehicle routing problem with deliveries and selective pickups (SVRPDSP). A vehicle departs loaded from the depot, visit every customer delivering a certain amount of goods according to their demand, and optionally pickup items from those customers, receiving a profit for each pickup realized. The vehicle has a limited capacity, which may turn impossible to attend all pickups, or make this unprofitable if it has to come back later in the customer after unloaded enough to fit the pickup demand. The objective is to find a minimal cost feasible route, the cost being the total travel costs minus the total revenue earned with pickups. Despite the many real applications, the literature is scarce. We propose an evolutionary algorithm whose crossover and mutation operators use data mining strategies to capture good characteristics from the parents and the population. Solutions are improved by a VNS algorithm during the process, and new solutions are introduced regularly to avoid premature convergence, using good constructive algorithms. The algorithm was tested with a benchmark of 68 instances, and the results compared to other publications. The results show the robustness of the method and 7 new solutions were found, including 2 new optimal solutions.

General hybrid column generation algorithm for crew scheduling problems using genetic algorithm

This paper describes a general hybrid column generation algorithm for crew scheduling problems, using genetic algorithm to speed up the generation of new columns, combined with an integer programming exact method to assure optimality. The subproblem of the column generation must generate a new feasible set of tasks to be assigned to a crew member. It is modeled as a shortest path with resource constraints problem in a graph, which virtually can be applied to all kinds of crew scheduling problems. The genetic algorithm is also general, and knowledge about specific problems may be incorporated. The hybrid algorithm is tested with instances from the literature and also with real instances, and the results show that the genetic algorithm is able to quickly generate most of the columns needed to solve the problem, while the exact method generates the last columns to find the optimal solution. The algorithm can also incorporate other kind of heuristics.

A multi-objective genetic algorithm with path relinking for the p-median problem

This paper considers the p-median problem that consists in finding p- locals from a set of m candidate locals to install facilities minimizing simultaneously two functions: the sum of the distances from each customer to its nearest facility and the sum of costs for opening facilities. Since this is a NP-Hard problem, heuristic algorithms are the most suitable for solving such a problem. To determine nondominated solutions, we propose a multi-objective genetic algorithm (MOGA) based on a nondominated sorting approach. The algorithm uses an efficient elitism strategy and an intensification operator based on the Path Relinking technique. To test the performance of the proposed MOGA, we develop a Mathematical Programming Algorithm, called eConstraint, that finds Pareto-optimal solutions by solving iteratively the mathematical model of the problem with additional constraints. The results show that the proposed approach is able to generate good approximations to the nondominated frontier of the bi-objective problem efficiently.

A combination of evolutionary algorithm, mathematical programming, and a new local search procedure for the just-in-time job-shop scheduling problem

This paper presents a combination of evolutionary algorithm and mathematical programming with an efficient local search procedure for a just-in-time job-shop scheduling problem (JITJSSP). Each job on the JITTSSP is composed by a sequence of operations, each operation having a specific machine where it must be scheduled and a due date when it should be completed. There is a tardiness cost if an operation is finished later than its due date and also an earliness cost if finished before. The objective is to find a feasible scheduling obeying precedence and machine constraints, minimizing the total earliness and tardiness costs. The experimental results with instances from the literature show the efficiency of the proposed hybrid method: it was able to improve the known upper bound for most of the instances tested, in very little computational time.

Hybrid approach for the Multiple Vehicle Routing Problem with Deliveries and Selective Pickups

We propose the Multiple Vehicle Routing Problem with Delivery and Selective Pickups (MVRPDSP) along with a mixed integer linear programming formulation and a hybrid cluster-first heuristic. We show that it is possible to divide the cluster-first heuristic in sub problems which can be solved exactly and their solutions combined in a semi-greedy way to generate good solutions. The results show that while the model is not able to find even a feasible solution for several instances the heuristic finds, for most cases, better or equal solutions in a matter of milliseconds, including 4 optimal solutions, what proves its efficiency. We also tested a combined approach, using the heuristic to generate an initial solution for the model, which improved even more the results of the heuristic.

Reactive GRASP with path relinking for solving parallel machines scheduling problem with resource-assignable sequence dependent setup times

This paper addresses an unrelated parallel machine problem with machine and job sequence dependent setup times. The objective function considered is a linear combination of the total completion time and the total number of resources assigned. Due to the combinatorial complexity of this problem, we propose an algorithm based on the GRASP metaheuristic, in which the basic parameter that defines the restrictiveness of the candidate list during the construction phase is self-adjusted according to the quality of the solutions previously found (reactive GRASP). The algorithm uses an intensification strategy based on the path relinking technique which consists in exploring paths between elite solutions found by GRASP. The results obtained by the proposed algorithm are compared with the best results available in the literature.

A GRASP with Path Relinking for the Single Machine Total Weighted Tardiness Problem

This study considers a single machine scheduling problem with the objective of minimizing the total weighted tardiness of the jobs. This problem is one of the most famous problems in single machine scheduling theory and it is NP-hard. In this paper, we propose a hybrid heuristic which combines GRASP with Path Relinking to find good quality solutions for the considered problem. The performance of the hybrid GRASP heuristic is tested over multiple benchmark problems from the OR-Library with up to 100 jobs and experimental results show that it is very competitive with the existing good performing algorithms.

Heuristic and Exact Approach for the Close Enough Ridematching Problem

The great number of vehicles in the streets is one of the biggest problems in big cities. Ridesharing, which has shown itself as a way to reduce the impact of this problem, is a subject widely discussed in the academic community nowadays. However, to the best of our knowledge, there is no paper in this subject including the characteristics we use in our work. In our approach, a person that offers a ride does not need to pass at the origin point of the person that request a ride but just at a point close enough of it. This way, we have an approach closer to what happens in practice. In this paper, we propose a MILP formulation to the problem and a heuristic. Then we propose a hybrid algorithm combining both approaches to reach good quality solutions. We show results for instances with different numbers of riders and drivers, using a map based on a real medium-size city.

Heuristic approaches to Double Vehicle Routing Problem with Multiple Stacks

The Double Routing Vehicle Problem with Multiple Stacks (DVRPMS) consists in a Double Traveling Salesman Problem with Multiple Stacks (DTSPMS) with multiple vehicles. Both problems appeared for the urgent need of optimizing intermodal transportation in the european context. Consolidated economies call for logistics studies, particularly on intermodal transportation, which consists of using different transport modes between a start point and an end point. In DVRPMS, the collecting region is usually connected to the delivery region by sea, and the cities that make up each of these regions are interconnected by land routes. In this paper, we propose three heuristics for the DVRPMS based on the Iterated Local Search (ILS), Simulated Annealing (SA) and Variable Neighborhood Descent (VND). The objective is to provide a route with near optimum cost in a timeframe that complies with the customers needs. The heuristics were applied in known instances and compared to the exact method. Both ILS and SA algorithms showed to be satisfactory for small instances and open to improvement for large instances. The SA algorithm managed to find solutions better than the best known upperbound. The VND algorithm served more as a guide to the quality of the initial solution than as a provider of a good final solution.

Multi-objective Variable Neighborhood Search algorithms for a just-in-time single machine scheduling problem

In this paper, we analyze the performance of two multi-objective algorithms based on the Variable Neighborhood Search (VNS). The first algorithm was proposed by Geiger and the second algorithm is proposed in this work. The algorithms are applied to solve the single machine scheduling problem with sequence dependent setup times and distinct due windows. We consider minimizing the total weighted earliness/tardiness and the total flow time criteria. The proposed multi-objective VNS algorithm generates good non-dominated solutions (an approximation of the Pareto-optimal solutions) in reasonable computation time. The performance of the algorithms is tested on a set of medium and large size instances. The computational results show that our approach is promising heuristic for multi-objective optimization.