Design Patterns

Real-world planning problems often require more than basic optimization. This section covers patterns for common scenarios.

Topics

• Real-Time Planning - Handle changes while the solver is running
• Continuous Planning - Rolling horizon and replanning strategies
• Repeated Planning - Batch optimization patterns

Real-Time Planning

Handle dynamic changes during solving:

from solverforge_legacy.solver import ProblemChange

class AddLessonChange(ProblemChange[Timetable]):
    def __init__(self, lesson: Lesson):
        self.lesson = lesson

    def do_change(self, working_solution: Timetable, score_director):
        # Add the new lesson to the working solution
        working_solution.lessons.append(self.lesson)
        score_director.after_entity_added(self.lesson)

# Apply change while solver is running
solver.add_problem_change(AddLessonChange(new_lesson))

Continuous Planning

For problems that span long time periods, use a rolling horizon:

1. Plan Window - Only optimize a subset of the timeline
2. Published Window - Lock decisions that are being executed
3. Draft Window - Future decisions that can still change

When to Use These Patterns