A general-purpose backtracking constraint satisfaction algorithm

A general-purpose backtracking constraint satisfaction algorithm.

Usage

A constraint satisfaction problem (CSP) can be used to model problems where a solution must be found to satisfy one or more constraints. Examples of such a problem include event scheduling, or map coloring.

A CSP framework is a set of variables that can be assigned within a range called a domain. For example, If we are cooking dinner for three guests , we would define them as variables that can be assigned a meal:

var doug = Person('Doug', dislikes: ['artichoke']);
var patrick = Person('Patrick', dislikes: ['bananas']);
var susan = Person('Susan', dislikes: ['broccoli']);

var variables = [doug, patrick, susan];

And the domains (the meal for each guest) could be modeled as a list strings, one of which will be assigned to each variable (guest).

var meals = ['artichoke', 'bananas', 'broccoli'];

var domains = {
  doug: meals,
  patrick: meals,
  susan: meals,
};

Finally, a constraint can be defined by subclassing of the Constraint class:

class AvoidDislikes extends Constraint<Person, String> {
  AvoidDislikes(super.variables);

  @override
  bool isSatisfied(Map<Person, String> assignment) {
    // ...
  }
}

To run a backtracking search of the solution space, instantiante the CSP class and call the backtrackingSearch function:

var csp = CSP<Person, String>(variables, domains);

csp.addConstraint(AvoidDislikes(variables));

var result = csp.backtrackingSearch();
print(result);

For complete examples, see the examples/ directory.

References

Classic Computer Science Problems in Java – David Kopec

GitHub

View Github