AI Glossary / AI and machine learning / Intermediate
Knowledge-based configuration
Definition
Knowledge-based configuration is the use of explicit domain knowledge to assemble a product, service or system from available components while satisfying requirements and compatibility constraints. It helps determine which combinations are valid and which choices remain possible.
What it means
Many offerings are configurable: computers, industrial equipment, insurance packages and software deployments can contain numerous interacting options. A simple list of available parts is insufficient because choosing one option may require another or exclude a third.
A knowledge-based configurator represents these relationships so it can guide a user through a feasible combination. It may explain why a choice is unavailable, identify missing components or suggest changes when requirements conflict.
Configuration is different from merely recommending a popular item. Its central responsibility is to respect the formalized constraints of the offering. It is also different from general design: configuration usually works within a defined catalog of components and relationships, although real applications can combine configuration with other design activities.
How it works
Domain experts describe components, attributes and constraints. A constraint might state that a selected motor requires a compatible power supply or that a package cannot include two mutually exclusive options. Customer requirements add further restrictions.
A reasoning engine searches for assignments that satisfy these conditions. Constraint satisfaction, rules, logic-based reasoning and related methods can support this process. As users make choices, propagation can narrow the remaining options without enumerating every complete combination.
When no solution exists, diagnosis or explanation methods can identify conflicting requirements or problematic knowledge-base rules. The system may offer a minimal set of changes, depending on its capabilities. Maintenance is essential: an outdated compatibility rule can produce an invalid order even when the reasoning engine correctly follows its stored knowledge.
Three practical examples
1. An equipment seller guides customers through a machine configuration, checking power, capacity and accessory compatibility.
2. A software provider assembles a deployment package that respects operating-system, database and feature dependencies.
3. A service designer checks whether a requested bundle of support hours, locations and response commitments can be delivered with the available service components.
Think of it this way
Think of assembling a kit with a knowledgeable guide who knows which pieces fit and what each choice requires. The limit is that the configurator only enforces knowledge that has been represented correctly; it cannot rely on the unspoken practical judgment an experienced engineer might apply.
Worked example
Suppose a toy catalog has 3 motors, 2 power supplies and 2 enclosures. Without constraints there are 3 × 2 × 2 = 12 combinations.
Assume motor M3 requires supply S2 and enclosure E2. For M1 and M2, all four supply-enclosure combinations remain valid, giving 8 configurations. M3 contributes only the one required combination, so 9 configurations satisfy these simplified rules.
If a customer requires M3 together with E1, there is no valid configuration under the stated knowledge. The useful response is to explain the conflict and identify a possible requirement change, not silently substitute a part. Real catalogs can contain many more interacting constraints.
Illustrative case study
A fictional packaging-machine supplier introduces an online configurator. Customers can complete an order, but engineers discover that one new accessory requires a controller version absent from the knowledge base.
The team adds the dependency, tests known valid and invalid configurations and assigns responsibility for updating rules when products change. It also adds an explanation that tells customers why the accessory requires a different controller.
The case illustrates that configuration quality depends on both reasoning and knowledge maintenance. A system can be logically consistent with its own rules while still producing a physically unsuitable offering if those rules are incomplete.
Common misconceptions and mistakes
- A valid configuration is automatically the best configuration. Feasibility and preference or cost optimization are separate questions.
- A configurator can infer every missing engineering rule. Domain knowledge must be supplied or learned through a justified process.
- A catalog update is only a content change. It can alter compatibility and require regression checks of the knowledge base.
Business use
Knowledge-based configuration can reduce invalid orders and make complex offerings easier to sell and support. Its value is strongest when compatibility knowledge is important and combinations are too numerous for simple manual checking.
Define which constraints are mandatory and which preferences can be negotiated. Keep explanations understandable to customers and staff. Maintain representative valid, invalid and conflicting requests as the catalog evolves. Connect configuration output to fulfillment carefully so that the selected components, quantities and versions remain consistent throughout the order process.
People also ask
Does configuration require machine learning?
No. Explicit constraints and logical reasoning can be sufficient.
What happens when requirements conflict?
A capable system can explain the conflict and suggest changes, but the exact behavior depends on its design.
Can a configurator optimize price?
Yes, if an optimization objective is added, while still respecting required constraints.
Check your understanding
A combination satisfies every stored rule but uses an accessory with an undocumented dependency. Is the reasoning engine necessarily malfunctioning?
Answer: No. The knowledge base may be incomplete even if the engine correctly reasons over the rules it was given.
Related terms
- Constraint satisfaction
- Knowledge base
- Diagnosis
- Optimization
- product configuration
- rule engine
- feature model
Learn next
Read constraint satisfaction and knowledge-base maintenance. Describe a small configurable product using variables, allowed values, compatibility rules and customer requirements, then test a deliberate conflict.
More in AI and machine learning
- Kernel methods for vector output - Kernel methods for vector output extend kernel learning to predict several output values together.
- Kernel Support Vector Machines - Kernel Support Vector Machines are SVMs equipped with a kernel function, letting them find non-linear boundaries by working implicitly in an enriched...
- Keypoints - Keypoints are specific, meaningful locations detected on an object or image, such as the corners of a document, the joints of a body or the landmarks of a...
- Knowledge-based recommender system - A knowledge-based recommender system suggests items by reasoning about explicit user requirements and knowledge of how items meet those requirements.
- Knowledge-based systems - Knowledge-based systems are computer systems that use explicitly represented knowledge to support reasoning, problem solving or decisions.
- Knowledge compilation - Knowledge compilation is the transformation of a knowledge representation into another form that supports selected reasoning queries more efficiently.
Sources
Felfernig, Hotz, Bagley and Tiihonen, Knowledge-Based Configuration: From Research to Business Cases (2014), author institution record: https://researchportal.helsinki.fi/en/publications/knowledge-based-configuration-from-research-to-business-cases/. Felfernig and colleagues, Consistency-Based Diagnosis of Configuration Knowledge Bases: https://www.sciencedirect.com/science/article/pii/S0004370203001176. The catalogue and supplier scenario are fictional. Source review: the book record confirms bibliographic details; direct access to the journal article failed. Relevant configuration and diagnosis definitions were checked in the authors’ earlier AAAI workshop paper: https://cdn.aaai.org/Workshops/1999/WS-99-05/WS99-05-008.pdf.
Previous: Knowledge-based agents Next: Knowledge-based recommender system