Are continuous domains always unbounded?

No. A finite interval such as [0, 3] is bounded but contains continuously many real values.

Does a computer store every real number exactly?

No. Numerical implementations use representations with limitations, even when the mathematical model uses real variables.

Can one problem mix continuous and discrete domains?

Yes. Different variables can represent divisible amounts, integer counts, and categorical choices within one model.

AI Glossary / Graphs and symbolic AI / Beginner

Continuous domains

Definition

Continuous domains are sets of possible variable values modeled as a continuum, such as all real numbers in an interval. In a constraint problem, a variable with a continuous domain can take intermediate values rather than only separate listed choices. Examples include modeled durations, distances, quantities, and physical positions.

What it means

A domain tells a model which values a variable may take. A continuous domain such as the interval from 0 to 10 includes 2, 2.5, 2.501, and indefinitely many other real values between its endpoints.

This differs from a discrete domain such as {0, 1, 2, 3}. It also differs from the integers, which are infinite but remain separated into individual values. Infinite does not automatically mean continuous.

The domain describes the mathematical model, not the precision of a measuring instrument or computer. A system may model time continuously while recording timestamps to the nearest second. Whether that approximation is suitable depends on the task and the meaning of its constraints.

How it works

Define each variable with units and an allowed region, then express restrictions through equations, inequalities, or other suitable relations. For instance, a duration may be nonnegative, and two durations may need to sum to a fixed total.

A solver reasons about regions of values rather than necessarily enumerating every value. Linear relationships can support specialized linear methods. Nonlinear relationships can introduce curved regions, disconnected feasible sets, or other difficulties requiring different techniques.

Some methods narrow interval bounds; others search or optimize using numerical calculations. The exact behavior depends on the solver and problem class. A continuous domain by itself does not guarantee an easy problem or an accurate numerical answer.

If implementation requires a finite grid, discretization changes the available choices. That can remove valid solutions or introduce rounding problems. The resulting assignments should be checked against the original requirements, with any tolerances clearly specified.

Three practical examples

1. A production model allocates divisible material quantities between processes, using real-valued amounts subject to capacity limits.

2. A robot model represents position coordinates continuously, even though sensors and controllers operate with finite resolution.

3. A scheduling model allows variable task durations in hours, while a later operational interface may restrict booking times to fixed minute increments.

Think of it this way

Think of a slider that can stop anywhere along a marked interval, compared with a selector offering a few numbered buttons. The slider represents the continuous model; the buttons represent discrete choices.

The analogy is only conceptual. A physical or digital slider has finite resolution, and some domain restrictions involve several variables together rather than one simple interval. A continuous model does not promise infinitely precise control of the real world.

Worked example

Suppose two divisible quantities x and y must total 7 units. Both are nonnegative, and x cannot exceed 3. The constraints are x + y = 7, x ≥ 0, y ≥ 0, and x ≤ 3.

Every real x in [0, 3] gives a feasible pair by setting y = 7 − x. Thus (2.5, 4.5) is valid, as are (0, 7) and (3, 4). The feasible set contains a continuum of points on a line segment.

If a separate requirement restricts both quantities to integers, only four pairs remain: (0, 7), (1, 6), (2, 5), and (3, 4). That is a different domain specification.

Now suppose the continuous problem also requires x = 2.5. It has the feasible pair (2.5, 4.5), while the integer version has no solution. Rounding x to 3 and leaving y at 4.5 violates the total, showing why numerical adjustments cannot be made independently without rechecking constraints.

These are exact illustrative calculations, not claims about a particular solver's numerical tolerance.

Illustrative case study

Fictional case: Harbour Studio models how to divide a batch of reusable material between two fabrication tasks. The planning model treats quantities as continuous, but the workshop actually issues material in fixed package sizes.

Before using the plan, a technician points out the packaging rule. The team changes the relevant decisions to integer package counts and keeps physical quantities as derived values. The revised model better matches execution. The case shows that choosing a domain is a practical modeling decision, not merely a mathematical convenience.

Common misconceptions and mistakes

Do not treat an unlimited set of whole numbers as continuous. The spacing between permitted integer values remains meaningful.

Do not assume every real quantity should be a continuous decision. Tickets, people, and indivisible components usually require discrete treatment even if averages can be fractional.

Avoid interpreting a solver's displayed decimal as exact physical precision. Representation, rounding, and feasibility tolerances need separate consideration.

Do not assume continuous means smooth. A model can have continuous variables together with nonsmooth functions or complicated restrictions. Likewise, an optimization objective is an additional feature; continuous feasibility alone does not specify a best solution.

Business use

Continuous domains support models for allocation, geometry, timing, and process control. Their usefulness depends on matching the decision's divisibility, units, and operational resolution.

Record which quantities can vary freely and which must occur in fixed increments. Where rounding is unavoidable, define an approved conversion and validation step. For mixed decisions, a model may combine continuous amounts with discrete choices, such as whether a facility is selected or how many packages are ordered.

Check your understanding

A variable may take any nonnegative integer. Is its domain continuous because it has infinitely many values?

Answer: No. It is an infinite discrete domain. Values such as 1.5 are still excluded, unlike an interval of nonnegative real numbers.

Related terms

Domain
Constraint satisfaction
Linear programming
Discretization
discrete domain
real-valued variable
mixed-integer programming
numerical tolerance

Learn next

Study how linear equations and inequalities define feasible regions. Plot the worked example as a line segment, then mark its integer points. Compare what happens when a grid is coarsened, especially when an equality requires a value between grid points.

Sources

Primary source: Russell and Norvig, Artificial Intelligence: A Modern Approach, constraint satisfaction chapter: https://aima.cs.berkeley.edu/2nd-ed/newchap05.pdf

This entry uses the constraint-modeling sense of continuous domains. The allocation example, rounding discussion, and fictional case are original teaching material.

Previous: Continuous analytics Next: Continuous feature