- Atlas
- Engineering
- Operations Research
Operations Research
Operations research (OR) is the discipline of using mathematics and systematic analysis to help people make better decisions about complex systems—how to route a fleet of trucks, how many beds a hospital should staff, where to place warehouses, which projects to fund.
17 frameworks, from 1937 to today · Industrial Engineering · Engineering
- Operational Research 1937-1960
- Systems Analysis and Operational Research Practice 1937-Present
- Monte Carlo method 1945-Present
- Linear Programming 1947-Present
- Systems Analysis 1948-1980
- Queueing theory 1950-Present
- Stochastic Modeling and Markov Decision Processes 1950-Present
- Dynamic Programming 1952-Present
- Network flow problem 1955-Present
- Optimization Under Uncertainty and Robust Optimization 1955-Present
- Stochastic Programming 1955-Present
- Combinatorial optimization 1958-Present
- Decision Analysis 1960-Present
- Discrete-Event Simulation 1960-Present
- Heuristics and Metaheuristics 1983-Present
- Robust Optimization 1998-Present
- Learning-Integrated Optimization 2010-Present
Overview
Operations research (OR) is the discipline of using mathematics and systematic analysis to help people make better decisions about complex systems—how to route a fleet of trucks, how many beds a hospital should staff, where to place warehouses, which projects to fund1. Its practitioners build models of a situation, solve them with algorithms, and use the answers to guide real choices1. What is at stake is not abstract elegance but the allocation of scarce resources: money, time, fuel, people, capacity. OR grew out of wartime necessity, matured into a mathematical science, and now reaches into logistics, health care, finance, energy, and public policy. Its history is a story of problems that resisted simple arithmetic and the ideas invented to crack them.
Wartime Origins and the Birth of a Discipline
The name and the field took shape during the Second World War, though its roots reach back further1. In Britain, the invention of radar in the late 1930s created a problem no single engineer could solve: how to site radar stations, how to filter their signals, and how to coordinate them with fighter squadrons. In 1937 the British physicist Robert Watson-Watt and colleagues had demonstrated radar's feasibility, and by 1940 a small group of scientists—physicists, physiologists, mathematicians—was working with the Royal Air Force on what they called "operational research": research on operations themselves, not on the equipment. They studied the size of convoys crossing the Atlantic, the depth settings of anti-submarine depth charges, and the scheduling of bomber missions. Their finding that larger convoys lost proportionally fewer ships to submarines, for instance, changed convoy policy.
The pattern was repeated across the Allied war effort. In the United States, the Navy and Army Air Forces assembled civilian scientists to analyze antisubmarine warfare, bombing accuracy, and logistics. The mathematician George Dantzig, then working for the U.S. Air Force, was later tasked with planning problems that would lead to his central contribution12. The war taught that teams of scientists from different backgrounds, working on real operations with data, could improve decisions that had previously been made by intuition and rank.
Linear Programming and the Simplex Method
The decisive mathematical breakthrough came just after the war. In 1947, Dantzig, working at the RAND Corporation and the Air Force, formulated the general linear programming problem: maximize or minimize a linear objective subject to linear constraints. The classic illustration is a diet problem—choose quantities of foods to meet nutritional requirements at minimum cost—but the same structure fit scheduling, blending, and transportation. Dantzig invented the simplex method, an algorithm that moves from one corner of the feasible region to another, improving the objective at each step, until no better corner exists2. It was fast, practical, and general.
Linear programming gave OR its first universal tool, and it spread quickly. The economist Tjalling Koopmans applied it to transportation and production planning, and Leonid Kantorovich in the Soviet Union had independently formulated related problems in 1939. The transportation problem—shipping goods from sources to destinations at least cost—became a standard model, and the Hungarian mathematician Harold Kuhn published an assignment algorithm in 1955. By the 1950s, linear programming was being taught in engineering and economics departments and used in oil refining, where blending crude oils into gasoline under quality constraints was a natural fit.
Networks, Integer Programming, and Combinatorial Structure
Many real problems are not smoothly divisible: you cannot ship half a truck or open two-thirds of a warehouse. Integer programming restricts some or all variables to whole numbers, and it proved far harder than linear programming. Ralph Gomory's cutting-plane method in the late 1950s gave a general algorithm, though early versions were slow3. The field of combinatorial optimization grew around such problems—scheduling, routing, packing—where the number of possible arrangements explodes4.
Network flow theory gave these problems a geometric and algebraic home5. In the 1950s, Lester Ford and Delbert Fulkerson developed the max-flow min-cut theorem, which states that the maximum amount of flow through a network equals the capacity of its narrowest cut. Their algorithms, and later refinements, made it possible to solve routing and distribution problems on graphs. The traveling salesman problem—find the shortest tour visiting a set of cities—became the emblem of combinatorial difficulty, and its study drove advances in exact methods and, later, heuristics6.
Uncertainty, Queues, and Stochastic Models
Real operations are full of randomness: arrivals at a service desk, machine breakdowns, demand fluctuations. Queuing theory addressed waiting lines7. The Danish engineer Agner Krarup Erlang had derived formulas for telephone exchange congestion as early as 1909, and his work was rediscovered and extended in the 1950s. The M/M/1 queue—Poisson arrivals, exponential service times, one server—became the textbook model, and networks of queues followed. Stochastic modeling more broadly, including Markov chains and Markov decision processes, gave OR a way to reason about systems that evolve probabilistically1. Monte Carlo methods, developed at Los Alamos in the 1940s by Stanislaw Ulam and John von Neumann, allowed analysts to simulate random processes numerically when closed-form solutions were impossible8.
Simulation and Systems Analysis
When models became too complex for analytical solution, discrete-event simulation stepped in. The idea is to represent a system as a sequence of events—arrivals, departures, failures—and run the clock forward, collecting statistics9. Early simulation languages appeared in the late 1950s and 1960s, and by the 1970s simulation was a standard tool for factory design, hospital planning, and military logistics. It let analysts test policies without disrupting real operations.
In parallel, systems analysis and the broader practice of OR matured into a profession. The RAND Corporation, founded in 1948, became a hub for applying OR to defense, budgeting, and policy. Decision analysis, developed largely by Howard Raiffa and Robert Schlaifer in the 1960s, brought subjective probability and utility theory into OR, giving decision-makers a formal way to weigh uncertain outcomes. The field's practitioners increasingly worked in teams, combining modeling with data collection and communication with managers.
Optimization Under Uncertainty and Robustness
Classical optimization assumed known data, but real data are uncertain. Stochastic programming, developed from the 1950s onward, optimized expected performance over scenarios. Robust optimization, formalized in the late 1990s and 2000s, instead sought solutions that perform acceptably under the worst plausible realization of uncertain parameters—a hedge against model error. These approaches changed how OR handled risk, moving from point estimates to ranges and distributions.
Heuristics, Metaheuristics, and Nature-Inspired Search
Exact algorithms guarantee optimality but can be too slow for large combinatorial problems. Heuristics—rules of thumb that find good but not provably optimal solutions—have always been part of OR. From the 1980s, metaheuristics generalized the idea: simulated annealing (inspired by cooling metals), genetic algorithms (inspired by evolution), tabu search, and ant colony optimization. These methods search vast solution spaces by combining exploration and exploitation. They are not guaranteed to find the best answer, but they often find excellent ones quickly, and they brought OR to problems—vehicle routing, scheduling, network design—that exact methods could not touch at realistic scale.
Learning-Integrated Optimization
The most recent turn integrates machine learning with optimization. Rather than assuming a model is given, analysts now use data to learn demand patterns, travel times, or failure rates, and then optimize against the learned model. This learning-integrated optimization—sometimes called prescriptive analytics—blurs the line between prediction and decision. It descends from the field's long tradition of combining statistical estimation with mathematical programming, but the scale of modern data and the flexibility of learning methods have opened new problems, from personalized pricing to real-time dispatch.
The Present
Operations research today is a broad, mature field. Linear and integer programming remain workhorses, solved by commercial and open-source solvers that handle millions of variables. Simulation, queuing theory, and decision analysis are standard in industry. Metaheuristics and learning-based methods address problems that resist exact solution. The field's questions are increasingly about integration: how to combine optimization with real-time data, how to make models robust to uncertainty and to human behavior, and how to deploy OR in settings—health care, sustainability, public policy—where objectives are multiple and contested. The wartime scientists who studied convoys would recognize the ambition, if not the tools.
References
- Operations research (introduction, Overview, Second World War, Historical origins), Wikipedia
- Simplex algorithm (History, introduction), Wikipedia
- Cutting-plane method (introduction, History), Wikipedia
- Combinatorial optimization, Wikipedia
- Network flow problem, Wikipedia
- Travelling salesman problem, Wikipedia
- Queueing theory, Wikipedia
- Monte Carlo method (introduction, History), Wikipedia
- Discrete-event simulation, Wikipedia
The frameworks of Operations Research
Each framework is a way of working the field has used: when it held, and the question it answers.
- 1937-1960
Operational Research
How can science improve real military and industrial operations?
- 1937-Present
Systems Analysis and Operational Research Practice
How should analysts model and advise on complex operations?
- 1945-Present
- 1947-Present
Linear Programming
How do we optimize a linear objective under linear constraints?
- 1948-1980
- 1950-Present
- 1950-Present
Stochastic Modeling and Markov Decision Processes
How should decisions be made in randomly evolving systems?
- 1952-Present
Dynamic Programming
How do we solve multistage sequential decision problems?
- 1955-Present
- 1955-Present
Optimization Under Uncertainty and Robust Optimization
How do we optimize when data are uncertain?
- 1955-Present
Stochastic Programming
How do we optimize expected outcomes over uncertain scenarios?
- 1958-Present
Combinatorial optimization
How do we find the best arrangement among discrete choices?
- 1960-Present
- 1960-Present
Discrete-Event Simulation
How can complex systems be simulated event by event?
- 1983-Present
Heuristics and Metaheuristics
How can good solutions be found quickly for hard problems?
- 1998-Present
- 2010-Present
Learning-Integrated Optimization
How can data and learning improve optimization decisions?
How the ideas connect
Each arrow runs from an earlier idea to the later one it shaped. The list below says what passed between them.
Systems Analysis and Operational Research Practice grew out of Operational Research
Wartime operational research teams in Britain and the US established the practice of scientists analyzing real operations with data; this practice was institutionalized after the war, notably at the RAND Corporation (founded 1948), as systems analysis and professional OR practice.
Carried by RAND Corporation.
Combinatorial optimization grew out of Linear Programming
Combinatorial optimization grew directly out of linear programming when variables were restricted to integers; Gomory's cutting-plane method in the late 1950s extended LP techniques to integer programs.
Carried by Ralph Gomory.
Robust Optimization grew out of Optimization Under Uncertainty and Robust Optimization
Robust optimization was formalized in the late 1990s as a branch of optimization under uncertainty, replacing probabilistic scenario models with worst-case parameter sets.
Linear Programming influenced by Operational Research
Dantzig's Air Force planning work, in the tradition of wartime operational research on logistics and scheduling, led him to formulate the general linear programming problem and the simplex method in 1947.
Carried by George Dantzig.
Systems Analysis grew out of Operational Research
Systems analysis grew out of wartime operational research's approach of applying scientific teams to military and policy decisions; RAND, founded in 1948, carried this forward into defense budgeting and policy analysis.
Carried by RAND Corporation.
Discrete-Event Simulation influenced by Monte Carlo method
Monte Carlo sampling of random processes provided the numerical technique that discrete-event simulation used to estimate statistics of stochastic systems when closed-form analysis failed.
Network flow problem influenced by Linear Programming
Linear programming's transportation model gave network flow theory its optimization framing; Ford and Fulkerson's max-flow min-cut work built on LP formulations of distribution problems.
Carried by Lester Ford; Delbert Fulkerson.
Optimization Under Uncertainty and Robust Optimization grew out of Linear Programming
Optimization under uncertainty extended linear programming by allowing uncertain data; stochastic programming from the 1950s optimized expected performance over scenarios within the LP framework.
Combinatorial optimization influenced by Network flow problem
Network flow theory gave combinatorial optimization its graph-theoretic models and algorithms; max-flow min-cut and shortest-path methods became core tools for routing and scheduling problems.
Carried by Lester Ford; Delbert Fulkerson.
Robust Optimization reacted against Stochastic Programming
Robust optimization arose partly as a response to stochastic programming's reliance on known probability distributions, instead hedging against worst-case realizations of uncertain parameters.
Heuristics and Metaheuristics influenced by Combinatorial optimization
Metaheuristics such as simulated annealing and genetic algorithms were developed from the 1980s to attack large combinatorial optimization problems that exact methods could not solve at realistic scale.
Discrete-Event Simulation influenced by Operational Research
When OR models of operations became too complex for analytical solution, analysts turned to discrete-event simulation; early simulation languages in the late 1950s and 1960s grew from this OR practice.
Heuristics and Metaheuristics influenced by Linear Programming
Heuristics and metaheuristics were developed largely to find good solutions to combinatorial problems that exact LP-based methods could not solve at realistic scale.
Learning-Integrated Optimization influenced by Linear Programming
Learning-integrated optimization uses mathematical programming, including linear and integer programming, as the optimization engine against models learned from data.
Combinatorial optimization influenced by Dynamic Programming
Dynamic programming provided exact solution methods for some combinatorial problems, such as sequencing and routing, complementing branch-and-bound and cutting-plane approaches.
Learning-Integrated Optimization influenced by Dynamic Programming
Dynamic programming's recursive decomposition of sequential decisions underlies reinforcement-learning approaches within learning-integrated optimization.
Learning-Integrated Optimization influenced by Combinatorial optimization
Learning-integrated optimization applies machine learning to combinatorial and other optimization problems, for instance learning heuristics or cost estimates to speed exact search.
Learning-Integrated Optimization influenced by Discrete-Event Simulation
Discrete-event simulation supplies synthetic data and environments used to train and evaluate learning-based decision policies in learning-integrated optimization.
Learning-Integrated Optimization influenced by Heuristics and Metaheuristics
Metaheuristic search ideas inform learning-integrated optimization, where learned models guide the exploration of large solution spaces.
Beyond the field
Where Operations Research's ideas came from in other fields, and what they led to there: the strongest 12 of 76 links.
Linear Programming influenced Least-cost ration formulation in Animal Nutrition
Least-cost ration formulation is the direct application of linear programming to feed mixing: the simplex method, published by Dantzig in 1947, was applied to least-cost feed formulas in the 1950s. The later framework is unthinkable without the earlier optimisation technique.
Carried by George Dantzig.
Linear Programming led to Integer and Mixed-Integer Programming in Combinatorial Optimization
Integer and mixed-integer programming was built directly on the linear-programming model developed in operations research, using LP relaxations as bounds and the simplex method as its core subroutine.
Carried by Ralph Gomory; George Dantzig.
Linear Programming influenced Least-Cost Linear Programming Formulation in Aquatic Nutrition
Operations-research linear programming, with its simplex algorithm and duality theory, is the mathematical basis of least-cost feed formulation. The method passed from operations research into agricultural and aquacultural ration optimisation.
Carried by George Dantzig.
Linear Programming led to Nonlinear Programming in Mathematical Optimization
Nonlinear programming grew directly out of linear programming: Kuhn and Tucker's 1951 paper generalised the simplex and duality framework to nonlinear constraints, and linear programming remains its special case.
Carried by Harold Kuhn; Albert Tucker.
Linear Programming influenced Branch and Bound in Combinatorial Optimization
The branch-and-bound procedure was devised within operations research to solve integer programs, relying on LP relaxation bounds from the simplex method at each node.
Carried by Ailsa Land; Alison Doig.
Systems Analysis influenced Systems Analysis in Political Science
Operations research, developed for military and industrial planning in the 1940s, supplied systems analysis with its core techniques: mathematical modeling of alternatives, cost-effectiveness comparison and optimization. The RAND-style policy systems analysis of the 1950s took these methods over directly.
Carried by RAND Corporation; Robert McNamara.
Dynamic Programming influenced Cost-based query optimization in Query Processing
Cost-based query optimizers used dynamic programming to enumerate join orders, directly adapting the technique from operations research. The System R optimizer's bottom-up dynamic programming over join subplans is the canonical case.
Carried by Patricia G. Selinger; System R (1979).
Dynamic Programming influenced Shortest-Path Label-Setting and Relaxation in Graph Algorithms
Dynamic programming, developed by Bellman from 1952, directly produced the Bellman-Ford shortest-path relaxation algorithm and the optimal-substructure view of shortest paths.
Carried by Richard Bellman.
Optimization Under Uncertainty and Robust Optimization led to Robust Optimization in Mathematical Optimization
Robust optimization grew directly out of the older optimization-under-uncertainty tradition, replacing probabilistic uncertainty with deterministic uncertainty sets. The 1990s formulation by Ben-Tal and Nemirovski refined that earlier line of work.
Carried by Aharon Ben-Tal; Arkadi Nemirovski.
Combinatorial optimization led to Facility location problem in Supply Chain Engineering
The facility location problem is a combinatorial optimisation problem, and its solution methods (branch and bound, Lagrangian relaxation, heuristics) come from that field. Hakimi's median problem and later uncapacitated facility location work are standard instances of combinatorial optimisation.
Carried by S. L. Hakimi.
Decision Analysis led to Robust Decision Making in Decision And Risk Analysis
RDM is a branch of operations-research decision analysis, using its formal alternatives, objectives and value trade-offs while shifting from optimising under known probabilities to stress-testing under deep uncertainty.
Carried by Robert Lempert.
Operational Research influenced by Queueing Theory in Service Operations
Queueing theory, founded by Erlang in 1909 for telephone traffic, supplied OR with one of its core mathematical models; postwar OR texts and practice took up waiting-line analysis for service and production systems.
Carried by A. K. Erlang.

Operations Research is a planet in Nousfarer
Its city has a tower for each of these 17 frameworks. Land, walk in, and climb one concept at a time.
Land on Operations ResearchRelated fields
Fields its ideas are linked to
- Mathematical Optimization 6 links
- Applied Mathematics 6 links
- Decision And Risk Analysis 6 links
- Systems Simulation 4 links
- Logistics 4 links
- Planning And Search 3 links
- Algorithms 3 links
- Combinatorial Optimization 3 links
- Supply Chain Engineering 3 links
- Managerial Decision Making 3 links
- Systems Engineering 3 links
- Query Processing 2 links
More in Industrial Engineering
- Human Factors
- Industrial Engineering
- Production Systems
- Quality Engineering
- Reliability Engineering
- Stochastic Modeling
- Supply Chain Engineering
- Systems Simulation
See where Operations Research sits on the Map of Human Knowledge