Skip to content
Back to search results

Multi-Objective Linear Programming (MOLP)

Introduced by
Steuer, R. E.; Charnes, A.; Cooper, W. W.
Year
1955–1986
Type
Mathematical optimization / vector optimization
Family
Simulation / optimization
Font size

What does it do?

Multi-Objective Linear Programming (MOLP) extends classical linear programming to handle several conflicting linear objective functions simultaneously over a feasible region defined by linear constraints.

Instead of a single optimal solution, MOLP produces a Pareto-efficient frontier from which a decision-maker selects a preferred trade-off. It is foundational to operations research and management science for resource allocation, planning, and design problems with competing goals.

The intuition behind the method

In ordinary linear programming you minimize one cost or maximize one profit, but real decisions almost always involve several competing goals — minimize cost while maximizing service level while minimizing emissions. MOLP acknowledges that no single solution is best on all objectives at once. It maps out the full frontier of non-dominated (Pareto-efficient) solutions where improving one objective would necessarily worsen another. The decision-maker then uses preference information — weights, aspiration levels, or interactive queries — to select a final solution from that frontier, making trade-offs explicit rather than hidden inside a composite score.

History

The theoretical foundations of multi-objective linear programming trace to the 1950s work of Charnes and Cooper on goal programming and Koopmans on efficient production. Steuer's 1986 textbook consolidated the field, systematizing the weighted-sum, epsilon-constraint, and interactive methods that remain standard today. MOLP is now a core topic in operations research curricula worldwide.

When to use it

Use MOLP when the problem has multiple competing linear objectives and linear constraints, data are continuous and precisely known (or can be treated as deterministic), and the decision-maker needs to understand trade-offs explicitly rather than accept a pre-aggregated score. Typical settings include production planning, portfolio allocation, transportation routing, and public resource budgeting. Do NOT use when objectives or constraints are non-linear (use multi-objective nonlinear programming or metaheuristics instead), when data are discrete or binary (use multi-objective integer programming), when the number of objectives exceeds five or six (the efficient frontier becomes computationally unwieldy), or when the decision-maker is unwilling to engage in iterative preference elicitation.

Limitations and considerations

  • Restricted to linear objective functions and linear constraints; real-world problems are often non-linear.
  • The efficient frontier can contain exponentially many extreme points in higher dimensions, making comprehensive enumeration impractical.
  • Requires precise numerical data; uncertainty or fuzziness in coefficients is not natively handled.
  • Interactive methods depend heavily on the decision-maker's availability and ability to articulate preferences consistently.

Source conditions do not replace independent scientific review or validation by an analysis engine.

Strengths

  • Produces the full Pareto-efficient frontier, giving decision-makers a complete picture of trade-offs.
  • Computationally tractable for moderate problem sizes via simplex-based solvers.
  • Guarantees mathematical optimality of every point on the efficient frontier.
  • Well-supported by mature software (CPLEX, Gurobi, open-source alternatives).
  • Trade-offs are made explicit rather than collapsed into an opaque composite index.

Applications

  • Production planning balancing cost, throughput, and resource utilization.
  • Portfolio optimization trading off return, risk, and liquidity.
  • Transportation and logistics network design with competing cost and service objectives.
  • Energy system planning minimizing cost while meeting emission and reliability targets.
  • Healthcare resource allocation balancing equity, efficiency, and budget constraints.
  • Environmental management optimizing economic and ecological performance simultaneously.

How it works

How it works

1

Formulate k linear objective functions f1(x) through fk(x) and a feasible region X defined by linear inequality and equality constraints (Ax <= b, x >= 0). Identify which objectives are to be minimized and which maximized. Confirm that all objective functions and constraints are linear in the decision variables.

Formula
\min_{x \in X} \mathbf{f}(x) = (f_1(x), f_2(x), \ldots, f_k(x))^T \text{ s.t. } Ax \leq b,\; x \geq 0
2

Generate the set of Pareto-efficient (non-dominated) solutions. Common approaches include the weighted-sum method (varying weights across the simplex), the epsilon-constraint method (optimize one objective while bounding the others), and the Chebyshev method. Each approach traces a portion or all of the efficient frontier.

3

Apply a scalarization technique to convert the vector problem into a sequence of single-objective LPs. In the weighted-sum approach, solve min sum(wi * fi(x)) for systematically varied weight vectors. Record each efficient extreme point of the objective space.

Formula
\min_{x \in X} \sum_{i=1}^{k} w_i f_i(x), \quad w_i \geq 0, \; \sum w_i = 1
4

Collect the efficient extreme points and the edges connecting them to characterize the Pareto frontier. Analyze the range of each objective across the frontier to understand the magnitude of trade-offs. Present the frontier to the decision-maker via tables or bi-objective plots.

5

Engage the decision-maker to select a preferred solution from the efficient set. Interactive methods (STEM, NIMBUS) alternate between presenting solutions and asking the decision-maker to express preference, converging on a final choice that best reflects their priorities.

Common pitfalls

  • Using fixed weights in the weighted-sum scalarization and reporting only one solution obscures the frontier and defeats the purpose of MOLP.
  • Confusing Pareto efficiency with optimality: every point on the frontier is efficient, but none is globally 'best' without preference information.
  • Neglecting to check whether the feasible region is bounded; an unbounded problem yields no finite efficient set.
  • Treating MOLP as a black box without engaging the decision-maker in selection leads to arbitrary final choices.
  • Applying MOLP to inherently integer problems without switching to multi-objective integer programming, producing solutions that violate integrality.

Frequently asked questions

How is MOLP different from goal programming?

Goal programming is a special case or variant: it converts each objective into an aspiration level and then minimizes deviations from those targets, yielding a single solution. MOLP in its general form seeks to characterize the full efficient frontier without pre-specifying target levels, giving the decision-maker more flexibility.

Can MOLP handle more than two objectives?

Yes, but the Pareto frontier becomes harder to visualize beyond two or three objectives and the number of efficient extreme points can grow exponentially. Interactive methods and representative-subset techniques are recommended for four or more objectives.

What if my data are uncertain?

Standard MOLP assumes deterministic coefficients. Uncertainty can be addressed by extending to stochastic multi-objective linear programming (chance constraints, scenario-based approaches) or robust multi-objective optimization, which are separate methods.

Is MOLP the same as multi-objective optimization?

MOLP is a special case of multi-objective optimization where all objectives and constraints are linear. Multi-objective optimization is the broader field that also includes non-linear, integer, and combinatorial variants.

Which scalarization method should I use?

The weighted-sum method is simple but misses non-convex parts of the frontier. The epsilon-constraint method can recover the full frontier even for non-convex regions and is generally preferred for thorough analysis. Interactive methods are best when decision-maker engagement is possible.

Related methods

Sources

  1. Steuer, R. E. (1986). Multiple Criteria Optimization: Theory, Computation, and Application. John Wiley & Sons, New York.

    ISBN: 9780471888468
    Copied.
  2. Chankong, V., Haimes, Y. Y. (1983). Multiobjective Decision Making: Theory and Methodology. North-Holland, New York.

    Open publication reference
    Copied.