Plan Space Search
Overview
There are always multiple ways to achieve a goal. Instead of committing to the first plan that seems workable:
- Generate diverse plans
- Evaluate against comprehensive criteria
- Select the optimal plan
This avoids local optima and premature commitment.
Goal
Generate multiple possible plans to achieve a goal, then search for the one that best satisfies optimization criteria.
Steps
Step 1: Define Goal and Constraints
Clearly state:
- What success looks like
- Hard constraints (must satisfy)
- Soft preferences (want to satisfy)
- Available resources
Output: Planning brief
Step 2: Generate Plan via Forward Planning
Start from current state. What actions lead toward goal? Build sequence to goal.
Output: Forward plan
Step 3: Generate Plan via Backward Planning
Start from goal state. What must happen right before? Work back to current state.
Output: Backward plan
Step 4: Generate Analogical Plans
What similar goals have been achieved? How were they achieved? Adapt those plans.
Output: Analogical plans
Step 5: Generate Radical Alternatives
What’s a completely different approach? Challenge assumptions:
- What if we did it in half the time?
- What if we had 10x budget?
- What if we had no budget?
- What if it had to be done by one person?
Output: Alternative plans
Step 6: Validate Constraint Satisfaction
For each plan, check:
- Does it satisfy all hard constraints?
- If not, can it be modified?
- If not modifiable, discard.
Output: Valid plans
Step 7: Estimate Resources
For each valid plan, estimate:
- Time required
- Money required
- People required
- Other resources
Output: Resource estimates
Step 8: Score Each Plan
Score on criteria:
- Achieves goal
- Complete
- Resource efficient
- Time efficient
- Handles uncertainty
- Has fallbacks
- Simple to execute
- Few dependencies
- Exit possible
Output: Scored plans
Step 9: Compare Plans
Create comparison table. Identify which plan wins on which criteria. Look for dominant plan.
Output: Plan comparison
Step 10: Sensitivity Analysis
How robust is the ranking?
- What if time was more important?
- What if resources were scarcer?
- What if uncertainty was higher?
Output: Sensitivity assessment
Step 11: Select and Refine
Select top plan. Can it incorporate strengths from other plans? Refine the plan with details.
Output: Final plan
Step 12: Document Contingencies
For the selected plan:
- What could go wrong?
- What are the triggers for Plan B?
- What parts of rejected plans serve as fallbacks?
Output: Contingency documentation
When to Use
- Planning a project
- Strategy development
- Resource allocation decisions
- Route/path finding
- Any situation with multiple approaches
Verification
- Multiple plan generation methods used
- Plans satisfy hard constraints
- Resource estimates are realistic
- Scoring is consistent
- Sensitivity analysis performed
- Contingencies documented