Facility Location

Facility Location
Name	Facility Location
Field	Operations research, Computer science, Discrete optimization
Introduced	20th century
Notable examples	Weber problem, p-median problem, k-means clustering

Facility Location is a class of optimization problems concerning the placement of service sites to minimize cost or maximize service quality for a set of demand points. The topic links combinatorial optimization traditions from George Dantzig and John von Neumann to algorithmic developments in Richard Karp's era, and has influenced applied work at institutions such as AT&T, IBM, and Google. Researchers at laboratories like Bell Labs, MIT Lincoln Laboratory, and Los Alamos National Laboratory have framed models that connect to classical problems studied in INFORMS and presented at conferences like STOC and SODA.

Introduction

The facility placement tradition traces roots to the Weber problem and the 1-median problem studied by scholars such as Thomas Simpson and developed further at Harvard University and Princeton University. Practical incarnations appeared in operations at FedEx, United Parcel Service, and UPS, and in planning efforts by agencies like the United States Postal Service. The subject unites techniques from Linear Programming pioneered at RAND Corporation, Integer Programming advanced at Cornell University, and approximation paradigms championed at Carnegie Mellon University and Stanford University.

Problem Variants

Variants include the p-median problem, the p-center problem, the uncapacitated facility location problem studied at Bell Labs, the capacitated facility location problem examined in projects at Deloitte and McKinsey & Company, and stochastic and dynamic versions explored by teams at University of California, Berkeley and Columbia University. Others include the k-center problem related to work at ETH Zurich, the k-means clustering variant popularized by researchers at AT&T Bell Laboratories, and hierarchical or multi-level formulations used in collaborations with World Bank projects and United Nations urban planning initiatives.

Mathematical Models and Formulations

Formulations deploy Integer programming models popularized by John von Neumann's linear optimization legacy, mixed-integer programming used in studies at IBM Research, and continuous relaxations reminiscent of techniques from George Dantzig and Philip Wolfe. Duality theory from Rockafellar informs Lagrangian relaxations applied by groups at Los Alamos National Laboratory. Metric assumptions tie to work on Fermat points and the Weber problem in classical geometry, while facility opening costs and assignment costs mirror cost structures analyzed in studies at McKinsey & Company and Boston Consulting Group.

Solution Methods and Algorithms

Exact methods include branch-and-bound techniques developed at Bell Laboratories and branch-and-cut frameworks implemented in solvers from Gurobi and CPLEX used by researchers at IBM and Microsoft Research. Approximation algorithms with provable ratios were advanced by Richard Karp, Vijay Vazirani at University of California, Berkeley, and Sanjeev Arora at Princeton University, with primal-dual schemas influenced by Jack Edmonds's matroid theory. Heuristics such as local search, greedy algorithms, and metaheuristics like simulated annealing and genetic algorithms have been employed in industrial studies at Siemens and General Electric. Online and streaming algorithms relate to work by Avi Wigderson and Noga Alon presented at FOCS and STOC.

Applications and Industry Examples

Real-world deployments occur in logistics at Amazon, distribution planning at Walmart, telecommunication hub placement by Verizon and AT&T, hospital location studies undertaken by teams at World Health Organization and Centers for Disease Control and Prevention, and emergency services optimization in projects with Federal Emergency Management Agency. Urban transit layout research involves collaborations with New York City, London Transport, and Transport for London, while retail network strategies appear in case studies by McKinsey & Company and Boston Consulting Group. Cloud infrastructure mapping and data center siting have been central to operations at Google, Microsoft Azure, and Amazon Web Services.

Complexity, Theory, and Performance Guarantees

Complexity results trace to Richard Karp's NP-completeness framework and subsequent hardness proofs by Umesh Vazirani and Vijay Vazirani; inapproximability bounds build on PCP theorems credited to Arora and Safra. Approximation guarantees such as constant-factor ratios for uncapacitated versions were established by researchers at Princeton University and Stanford University, while logarithmic bounds appear in work by David Shmoys and collaborators at MIT. Theoretical connections link to metric embeddings studied at UC Berkeley and to submodular optimization explored by G. L. Nemhauser at University of Maryland.

Extensions and Related Problems

Extensions include facility location with capacities, multi-echelon network design investigated by Dantzig School collaborators, facility location under uncertainty examined in studies at INSEAD, and competitive facility location or location games pursued in economic studies by Harvard University and London School of Economics. Related problems encompass network design like the Steiner tree problem studied at ETH Zurich, clustering formulations such as k-means and k-median used in machine learning at Google Research and DeepMind, and service facility scheduling problems integrated in work at Bell Labs and AT&T Bell Laboratories.

Category:Operations research