Travelling Salesman Problem

Travelling Salesman Problem
Xypron · Public domain · source
Name	Travelling Salesman Problem
Caption	Example of a route in a weighted complete graph
Field	Theoretical Computer Science
Introduced	1930s
Solved by	Exact algorithms, approximation heuristics
Complexity	NP-hard, NP-complete (decision version)

Travelling Salesman Problem

The Travelling Salesman Problem is a classical combinatorial optimization challenge that asks for the shortest possible route through a set of cities visiting each exactly once and returning to the start. Originating in the early 20th century, the problem has attracted contributions from figures and institutions across United States and United Kingdom mathematics and operations research communities, and influenced work at Bell Labs, IBM, MIT, Princeton University, and Stanford University. Research on the problem intersects with results from Stephen Cook-related complexity theory, techniques developed at Los Alamos National Laboratory, and computational studies by teams at Microsoft Research and Google.

Problem statement

Formally, given a finite set of vertices representing cities and weighted edges representing pairwise distances, the goal is to find a Hamiltonian circuit of minimum total weight in a complete weighted graph. Early computational investigations were performed by practitioners at University of California, Berkeley, scholars influenced by methods from John von Neumann and the Institute for Advanced Study, and by operations researchers affiliated with RAND Corporation and Cornell University. Variants and formalizations appeared in the literature connected to work by researchers associated with Bellman-type dynamic programming and development at Wright-Patterson Air Force Base logistical studies. Practical formulations often encode the problem as an integer program solved by solvers originating from teams at IBM Research and CERN.

Complexity and computational hardness

The decision form of the problem—whether a tour of length at most a bound exists—is NP-complete, a classification shaped by foundational theorems from Stephen Cook and Leonid Levin and by reductions studied at Princeton University and University of Chicago. The optimization version is NP-hard, linking it to landmark complexity results explored by researchers at Bell Labs, Harvard University, Yale University, and Columbia University. Hardness proofs draw on techniques developed in the broader NP-completeness program influenced by work at MIT and Stanford University and relate to structural complexity research from groups at Carnegie Mellon University and ETH Zurich. Deep connections exist between the problem and intractability results from Karp's list of NP-complete problems and subsequent refinements by scholars at University of Edinburgh and Rutgers University.

Exact algorithms

Exact solution methods include exhaustive search enhanced by branch-and-bound and cutting-plane methods, which were advanced by research groups at IBM and Bell Labs. The Held–Karp dynamic programming algorithm, with roots in combinatorial optimization studies influenced by Richard Bellman and collaborators at RAND Corporation, runs in exponential time but substantially reduces brute-force complexity. Branch-and-cut implementations were refined in computational experiments led by teams at Stanford University, INRIA, ETH Zurich, and University of Waterloo and incorporated into optimization systems developed by CERN and Tata Consultancy Services research groups. Exact solvers have leveraged parallel architectures studied at Argonne National Laboratory and Lawrence Livermore National Laboratory and benefited from integer programming advances by academics at University of Pennsylvania and University of Michigan.

Approximation and heuristics

Because exact methods scale poorly, approximation algorithms and heuristics are widely used. Christofides’ algorithm, derived from approximation theory advanced in collaboration with European groups including INRIA and researchers influenced by workshops at École Polytechnique, guarantees a 3/2-approximation for metric instances. Local search and metaheuristics—such as 2-opt, 3-opt, simulated annealing, tabu search, genetic algorithms, and ant colony optimization—were popularized by researchers at University of Toronto, Georgia Institute of Technology, University of Illinois at Urbana–Champaign, and teams inspired by work at Delft University of Technology and University of Bologna. Heuristic frameworks have been implemented in industrial tools produced by companies like IBM, SAP, and research labs at Microsoft Research and Amazon Web Services to tackle large-scale instances encountered in logistics and routing.

Special cases and variants

Numerous variants have spawned distinct literatures. The Euclidean case, where points lie in the plane, was analyzed by researchers at University of Cambridge, Princeton University, and ETH Zurich and admits polynomial-time approximation schemes under geometric constraints. The asymmetric version, studied by teams at Kyoto University and Tokyo Institute of Technology, allows directed edge costs and appears in scheduling contexts explored at University of Tokyo. Prize-collecting, time-window, vehicle-routing, and multiple-traveller variants have been developed for applications by groups at Massachusetts Institute of Technology, Columbia University, University of California, Los Angeles, and University of Southern California. The Graphical TSP, Bottleneck TSP, and Hamiltonian path restrictions link to combinatorial results from faculties at University of Oxford and University of Cambridge.

Applications and practical importance

The problem underpins routing and logistics solutions used by corporations like UPS and FedEx and has informed planning tools at municipal agencies such as the New York City Department of Transportation and the Transport for London network studies. It plays a central role in genome sequencing research at institutions like Wellcome Sanger Institute and in robotic path planning investigated at Carnegie Mellon University and ETH Zurich. Applications extend to manufacturing and circuit board drilling practices refined by engineers at Siemens and Bosch, and to astronomical survey scheduling employed by projects at European Southern Observatory and NASA Jet Propulsion Laboratory. The Travelling Salesman Problem remains a testbed for algorithmic innovation across universities and industry labs including MIT, Stanford University, Google DeepMind, and Facebook AI Research.

Category:Combinatorial optimization