The directional p-median problem: Definition,complexity, and algorithms

An instance of a p-median problem gives n demand points. The objective is to locate p supply points in order to minimize the total distance of the demand points to their nearest supply point. p-Median is polynomially solvable in one dimension but NP-hard in two or more dimensions, when either the Euclidean or the rectilinear distance measure is used. In this paper, we treat the p-median problem under a new distance measure, the directional rectilinear distance, which requires the assigned supply point for a given demand point to lie above and to the right of it. In a previous work, we showed that the directional p-median problem is polynomially solvable in one dimension; we give here an improved solution through reformulating the problem as a special case of the constrained shortest path problem. We have previously proven that the problem is NP-complete in two or more dimensions; we present here an efficient heuristic to solve it. Compared to the robust Teitz and Bart heuristic, our heuristic enjoys substantial speedup while sacrificing little in terms of solution quality, making it an ideal choice for real-world applications with thousands of demand points.     

The gravity p-median model

In this paper we propose a new model for the p-median problem. In the standard p-median problem it is assumed that each demand point is served by the closest facility. In many situations (for example, when demand points are communities of customers and each customer makes his own selection of the facility) demand is divided among the facilities. Each customer selects a facility which is not necessarily the closest one. In the gravity p-median problem it is assumed that customers divide their patronage among the facilities with the probability that a customer patronizes a facility being proportional to the attractiveness of that facility and to a decreasing utility function of the distance to the facility.     

Comparative error bound theory for three location models: continuous demand versus discrete demand

We develop a unified error bound theory to compare a given p-median, p-center or covering location model with continuously distributed demand points in R ⁿ to a corresponding given original model of the same type having a finite collection of demand points in R ⁿ . We give ways to construct either a continuous or finite demand point model from the other model and also control the error bound. Our work uses Voronoi tilings extensively, and is related to earlier error bound theory for aggregating finitely many demand points.     

An adaptive multiphase approach for large unconditional and conditional p-median problems

A multiphase approach that incorporates demand points aggregation, Variable Neighbourhood Search (VNS) and an exact method is proposed for the solution of large-scale unconditional and conditional p-median problems. The method consists of four phases. In the first phase several aggregated problems are solved with a “Local Search with Shaking” procedure to generate promising facility sites which are then used to solve a reduced problem in Phase 2 using VNS or an exact method. The new solution is then fed into an iterative learning process which tackles the aggregated problem (Phase 3). Phase 4 is a post optimisation phase applied to the original (disaggregated) problem. For the p-median problem, the method is tested on three types of datasets which consist of up to 89,600 demand points. The first two datasets are the BIRCH and the TSP datasets whereas the third is our newly geometrically constructed dataset that has guaranteed optimal solutions. The computational experiments show that the proposed approach produces very competitive results. The proposed approach is also adapted to cater for the conditional p-median problem with interesting results.     

Approximate solution of the <Emphasis Type="Italic">p</Emphasis>-median minimization problem

A version of the facility location problem (the well-known p-median minimization problem) and its generalization—the problem of minimizing a supermodular set function—is studied. These problems are NP-hard, and they are approximately solved by a gradient algorithm that is a discrete analog of the steepest descent algorithm. A priori bounds on the worst-case behavior of the gradient algorithm for the problems under consideration are obtained. As a consequence, a bound on the performance guarantee of the gradient algorithm for the p-median minimization problem in terms of the production and transportation cost matrix is obtained.     

Does Euclidean distance work well when the p-median model is applied in rural areas?

The p-median model is used to locate P centers to serve a geographically distributed population. A cornerstone of such a model is the measure of distance between a service center and demand points, i.e. the location of the population (customers, pupils, patients, and so on). Evidence supports the current practice of using Euclidean distance. However, we find that the location of multiple hospitals in a rural region of Sweden with a non-symmetrically distributed population is quite sensitive to distance measure, and somewhat sensitive to spatial aggregation of demand points.     

Marginal analysis for the fuzzy p-median problem

The solutions to the fuzzy p-median problem make it possible to leave part of the demand uncovered in order to obtain significant reductions in costs. Moreover, the fuzzy formulation provides the decision-maker with many flexible solutions that he or she may prefer to the classical crisp solution. We introduce some marginal analysis techniques to study how solutions depend on membership functions. Taking into account the internal structure of the problem, we propose a practical criterion to fix the tolerances for the uncovered demand, which happens to be the most sensitive aspect of the fuzzy p-median.     

Minmax regret location–allocation problem on a network under uncertainty

We consider a robust location–allocation problem with uncertainty in demand coefficients. Specifically, for each demand point, only an interval estimate of its demand is known and we consider the problem of determining where to locate a new service when a given fraction of these demand points must be served by the utility. The optimal solution of this problem is determined by the “minimax regret” location, i.e., the point that minimizes the worst-case loss in the objective function that may occur because a decision is made without knowing which state of nature will take place. For the case where the demand points are vertices of a network we show that the robust location–allocation problem can be solved in O(min{p, n − p}n³m) time, where n is the number of demand points, p (p < n) is the fixed number of demand points that must be served by the new service and m is the number of edges of the network.     

On the choice of aggregation points for continuousp-median problems: A case for the gravity centre

In large scale location-allocation studies it is necessary to use data-aggregation in order to obtain solvable models. A detailed analysis is given of the errors induced by this aggregation in the evaluation of thep-median objective function. Then it is studied how to choose the points at which to aggregate given groups of demand points so as to minimise this aggregation error. Forp-median problems with euclidean distances, arguments are given in favour of the centre of gravity of the groups. These arguments turn out to be much stronger for rectangular distance. Aggregating at the centroid also leads to much higher precision bounds on the errors for rectangular distance. Some numerical results are obtained validating the theoretical developments. This research was partially done while the author was on visit at the Laboratoire d’Analyse Appliquée et Optimisation at the Université de Bourgogne, Dijon, France. Thanks to E. Carrizosa, B. Rayco and four anonymous referees for many thoughtful remarks.     

Optimal algorithms for the α-neighbor p-center problem

Assigning multiple service facilities to demand points is important when demand points are required to withstand service facility failures. Such failures may result from a multitude of causes, ranging from technical difficulties to natural disasters. The α-neighbor p-center problem deals with locating p service facilities. Each demand point is assigned to its nearest α service facilities, thus it is able to withstand up to α − 1 service facility failures. The objective is to minimize the maximum distance between a demand point and its αth nearest service facility. We present two optimal algorithms for both the continuous and discrete α-neighbor p-center problem. We present experimental results comparing the performance of the two optimal algorithms for α = 2. We also present experimental results showing the performance of the relaxation algorithm for α = 1, 2, 3.     

An exact solution approach for the interdiction median problem with fortification

Systematic approaches to security investment decisions are crucial for improved homeland security. We present an optimization modeling approach for allocating protection resources among a system of facilities so that the disruptive effects of possible intentional attacks to the system are minimized. This paper is based upon the p-median service protocol for an operating set of p facilities. The primary objective is to identify the subset of q facilities which, when fortified, provides the best protection against the worst-case loss of r non-fortified facilities. This problem, known as the r-interdiction median problem with fortification (IMF), was first formulated as a mixed-integer program by Church and Scaparra [R.L. Church, M.P. Scaparra, Protecting critical assets: The r-interdiction median problem with fortification, Geographical Analysis 39 (2007) 129–146]. In this paper, we reformulate the IMF as a maximal covering problem with precedence constraints, which is amenable to a new solution approach. This new approach produces good approximations to the best fortification strategies. Furthermore, it provides upper and lower bounds that can be used to reduce the size of the original model. The reduced model can readily be solved to optimality by general-purpose MIP solvers. Computational results on two geographical data sets with different structural characteristics show the effectiveness of the proposed methodology for solving IMF instances of considerable size.     

Approximation in p-Norm of Univariate Concave Functions

We derive worst-case bounds, with respect to the L ^p norm, on the error achieved by algorithms aimed at approximating a concave function of a single variable, through the evaluation of the function and its subgradient at a fixed number of points to be determined. We prove that, for p larger than 1, adaptive algorithms outperform passive ones. Next, for the uniform norm, we propose an improvement of the Sandwich algorithm, based on a dynamic programming formulation of the problem.     

A compelling argument for the gravity p-median model

The p-median model is used to locate P facilities to serve a geographically distributed population. Conventionally, it is assumed that the population always travels to the nearest facility.  and  re-estate three arguments on why this assumption might be incorrect, and they introduce the gravity p-median model to relax the assumption. We favor the gravity p-median model, but we note that in an applied setting, the three arguments are incomplete. In this communication, we point at the existence of a fourth compelling argument for the gravity p-median model.     

Demand Point Aggregation for Planar Covering Location Models

The covering location problem seeks the minimum number of facilities such that each demand point is within some given radius of its nearest facility. Such a model finds application mostly in locating emergency types of facilities. Since the problem is NP-hard in the plane, a common practice is to aggregate the demand points in order to reduce the computational burden. Aggregation makes the size of the problem more manageable but also introduces error. Identifying and controlling the magnitude of the error is the subject of this study. We suggest several aggregation methods with a priori error bounds, and conduct experiments to compare their performance. We find that the manner by which infeasibility is measured greatly affects the best choice of an aggregation method.     

Probabilistic analysis of some euclidean clustering problems

We are given n points distributed randomly in a compact region D of R^m. We consider various optimisation problems associated with partitioning this set of points into k subsets. For each problem we demonstrate lower bounds which are satisfied with high probability. For the case where D is a hypercube we use a partitioning technique to give deterministic upper bounds and to construct algorithms which with high probability can be made arbitrarily accurate in polynomial time for a given required accuracy.     

Branch and Bound Algorithm for the <Emphasis Type="Italic">p</Emphasis>-Median Transportation Problem

The p-median transportation problem is to determine an optimal solution to a transportation problem having an additional constraint restricting the number of active supply points. The model is discussed as an example of a public sector location/allocation problem. A branch and bound procedure is proposed to solve the problem. Lagrangian relaxation is used to provide lower bounds. Computational results are given.     

The maximal-dispersion problem

In this paper, we consider the problem of selecting p pointsfrom m points, so that the p points are maximally dispersedwith respect to a specified metric. Two heuristics are studiedin terms of worst-case analysis, and a mathematical programmeis given, whose objective function is concave, together withan algorithm and error bounds on the loss of optimality arisingfrom early termination.     

The p-median problem: A survey of metaheuristic approaches

The p-median problem is one of the basic models in discrete location theory. As with most location problems, it is classified as NP-hard, and so, heuristic methods are usually used to solve it. Metaheuristics are frameworks for building heuristics. In this survey, we examine the p-median, with the aim of providing an overview on advances in solving it using recent procedures based on metaheuristic rules.     

Local networks for computer communications: A. WEST and P. JANSON (Eds.) Proceedings of the IFIP Working Group 6.4, International Workshop on Local Networks,Zürich,Switzerland, August 27–29, 1980, North-Holland,Amsterdam, 1981, xiii + 470 pages,Dfl.125.00

In this paper we present two lower bounds for the p-median problem, the problem of locating p facilities (medians) on a network. These bounds are based on two separate lagrangean relaxations of a zero-one formulation of the problem with subgradient optimisation being used to maximise these bounds. Penalty tests based on these lower bounds and a heuristically determined upper bound to the problem are developed and shown to result in a large reduction in problem size. The incorporation of the lower bounds and the penalty tests into a tree search procedure is described and computational results are given for problems with an arbitrary number of medians and having up to 200 vertices. A comparison is also made between these algorithms and the dual-based algorithm of Erlenkotter.     

Solving the p-Median Problem with a Semi-Lagrangian Relaxation

Lagrangian relaxation is commonly used in combinatorial optimization to generate lower bounds for a minimization problem. We study a modified Lagrangian relaxation which generates an optimal integer solution. We call it semi-Lagrangian relaxation and illustrate its practical value by solving large-scale instances of the p-median problem. This work was partially supported by the Fonds National Suisse de la Recherche Scientifique, grant 12-57093.99 and the Spanish government, MCYT subsidy dpi2002-03330.