Q1. Scenario: A warehouse robot must pick and place boxes. Use STRIPS planning: define the initial state, goal state, and actions (with preconditions and effects).
Initial: at(Robot, dock), box_on_pallet(A). Goal: box_on_shelf(A). Action pick(Box): preconditions at(Robot, BoxLocation), not holding(Box). Effects: holding(Box), not box_on_pallet(Box). Action place(Shelf): preconditions: holding(Box), at(Robot, Shelf). Effects: box_on_shelf(Box), not holding(Box).
Q2. Scenario: A personal assistant must plan a sequence: pick up coffee, then go to office, then attend meeting. Represent this as a partial-order plan. Explain why partial-order planning is more flexible than total-order.
Partial-order plan doesn't fully specify order of actions until necessary. For example, pick coffee and go to office can be in any order relative to each other, but both must occur before meeting. This reduces search and allows parallelism.
Q3. Scenario: In logistics, a planner must schedule deliveries under resource constraints (e.g., only 2 trucks). What type of planning is this? How does it differ from classical planning?
This is resource-constrained planning (or temporal planning). Classical planning ignores resources (trucks are unlimited). Resource planning tracks usage and availability, requiring additional constraints (e.g., not exceed 2 trucks).
Q4. Scenario: A robot in a dynamic environment where doors can be locked/unlocked by others. How does conditional planning differ from classical planning?
Conditional planning handles uncertainty by creating branches (if locked, then unlock; else proceed). Classical planning assumes deterministic actions and full observability. Conditional plans include sensing actions.
Q5. Scenario: Use hierarchical task network (HTN) planning for a vacation planning system. Give a high-level task and decompose it into subtasks.
High-level: PlanVacation. Decompose into: DecideDestination, BookFlight, BookHotel, PackItems. Each further decomposes (e.g., PackItems -> clothes, toiletries). HTN uses domain knowledge to constrain search.