Artificial Intelligence

Lecturer 12 - Planning

School of Information and Communication

Technology - HUST
Outline
• Planning problem
• State-space search
• Partial-order planning
• Planning graphs
• Planning with propositional logic

2
 Search vs. planning
• Consider the task get milk, bananas, and a cordless drill
• Standard search algorithms seem to fail miserably:

• After-the-fact heuristic/goal test inadequate

3
 Planning problem
• Planning is the task of determining a sequence of actions
that will achieve a goal.
• Domain independent heuristics and strategies must be
based on a domain independent representation
• General planning algorithms require a way to represent states,
actions and goals
• STRIPS, ADL, PDDL are languages based on propositional or
first-order logic
• Classical planning environment: fully observable,
deterministic, finite, static and discrete.

4
Additional complexities
• Because the world is …
• Dynamic
• Stochastic
• Partially observable
• And because actions
• take time
• have continuous effects

5
AI Planning background
• Focus on classical planning; assume none of the
above

• Deterministic, static, fully observable

• “Basic”
• Most of the recent progress
• Ideas often also useful for more complex problems

6
Problem Representation
• State
• What is true about the (hypothesized) world?
• Goal
• What must be true in the final state of the world?
• Actions
• What can be done to change the world?
• Preconditions and effects

• We’ll represent all these as logical predicates

7
STRIPS operators
• Tidily arranged actions descriptions, restricted language

• Action: Buy(x)
• Precondition: At(p); Sells(p; x)
• Effect: Have(x)

• [Note: this abstracts away many important details!]

• Restricted language  efficient algorithm

• Precondition: conjunction of positive literals
• Effect: conjunction of literals
• A complete set of STRIPS operators can be translated into a set of
successor-state axioms

8
 Example: blocks world
• On(b, x ): block b is on x, x is another block or Table. Initial
• MoveToTable(b, x): move block b from the top of x to Table
• Move(b, x, y): move block b from the top of x to the top of y C
• Clear(x): nothing is on x A B

◼ Action (Move (b, x, y) ,

❑ PRECOND: On(b, x)  Clear(b)  Clear(y),
Goal
❑ EFFECT: On(b, y)  Clear(x)  On(b, x)  Clear(y)) .
A
◼ Action(MoveToTable(b, x) ,
B
❑ PRECOND: On(b, x)  Clear(b),

❑ EFFECT: On(b, Table)  Clear(x)  On(b, x )) . C

◼ Initial state: On(A,Table), On(C,A), On(B,Table), Clear(B), Clear(C)

◼ Goal: On(A,B), On(B,C)

9
Planning with state-space search
• Both forward and backward search possible
• Progression planners
• forward state-space search
• consider the effect of all possible actions in a given state
• Regression planners
• backward state-space search
• Determine what must have been true in the previous state in
order to achieve the current state

10
Progression and regression
initial state

goal

11
Progression algorithm
• Formulation as state-space search problem:
• Initial state and goal test: obvious
• Successor function: generate from applicable actions
• Step cost = each action costs 1
• Any complete graph search algorithm is a complete planning
algorithm.
• E.g. A*
• Inherently inefficient:
• (1) irrelevant actions lead to very broad search tree
• (2) good heuristic required for efficient search

12
Forward Search Methods:
can use A* with some h and g

13
Regression algorithm
• How to determine predecessors?
• What are the states from which applying a given action leads to the goal?
Goal state = At(C1, B)  At(C2, B)  …  At(C20, B)
Relevant action for first conjunct: Unload(C1,p,B)
Works only if pre-conditions are satisfied.
Previous state= In(C1, p)  At(p, B)  At(C2, B)  …  At(C20, B)
Subgoal At(C1,B) should not be present in this state.
• Actions must not undo desired literals (consistent)
• Main advantage: only relevant actions are considered.
• Often much lower branching factor than forward search.

14
Regression algorithm
• General process for predecessor construction
• Give a goal description G
• Let A be an action that is relevant and consistent
• The predecessors are as follows:
• Any positive effects of A that appear in G are deleted.
• Each precondition literal of A is added , unless it already
appears.
• Any standard search algorithm can be added to perform the search.
• Termination when predecessor satisfied by initial state.
• In FO case, satisfaction might require a substitution.

15
 Backward search methods
Regressing a
ground
operator

16
Regressing an ungrounded
operator

17
Example of Backward Search
Heuristics for state-space search
• Use relax problem idea to get lower bounds on least number
of actions to the goal.
• Remove all or some preconditions
• Subgoal independence: the cost of solving a set of subgoals
equal the sum cost of solving each one independently.
• Can be pesimistic (interacting subplans)
• Can be optimistic (negative effects)
• Simple: number of unsatisfied subgoals.
• Various ideas related to removing negative effects or
positive effects.

19
Partial order planning
• Least commitment planning
• Nonlinear planning
• Search in the space of partial plans
• A state is a partial incomplete partially ordered plan
• Operators transform plans to other plans by:
• Adding steps
• Reordering
• Grounding variables
• SNLP: Systematic Nonlinear Planning (McAllester and
Rosenblitt 1991)
• NONLIN (Tate 1977)

20
A partial order plan for putting shoes
and socks

21
Partial-order planning
• Partially ordered collection of steps with
• Start step has the initial state description as its effect
• Finish step has the goal description as its precondition causal
links from outcome of one step to precondition of another
temporal ordering between pairs of steps
• Open condition = precondition of a step not yet causally
linked
• A plan is complete iff every precondition is achieved
• A precondition is achieved iff it is the effect of an earlier
step and no possibly intervening step undoes it

22
Example
Example
Example
Planning process
• Operators on partial plans:
• add a link from an existing action to an open condition
• add a step to fulfill an open condition
• order one step wrt another to remove possible conflicts

• Gradually move from incomplete/vague plans to

complete, correct plans

• Backtrack if an open condition is unachievable or if a

conflict is unresolvable

26
 POP algorithm sketch
function POP(initial, goal, operators) returns plan
plan  Make-Minimal-Plan(initial, goal)
loop do
if Solution?( plan) then return plan
Sneed , c  Select-Subgoal( plan)
Choose-Operator( plan, operators,Sneed, c)
Resolve-Threats( plan)
end

function Select-Subgoal( plan) returns Sneed , c

pick a plan step Sneed from Steps( plan)
with a precondition c that has not been achieved
return Sneed , c

27
POP algorithm (con’t)
procedure Choose-Operator(plan, operators,Sneed, c)
choose a step Sadd from operators or Steps( plan) that has c as as an effect
if there is no such step then fail
add the causal link Sadd →c Sneed to Links( plan)
add the ordering constraint Sadd < Sneed to Orderings( plan)
if Sadd is a newly added step from operators then
add Sadd to Steps( plan)
add Start < Sadd < Finish to Orderings( plan)

procedure Resolve-Threats(plan)
for each Sthreat that threatens a link Si →c Sj in Links( plan) do
choose either
Demotion: Add Sthreat < Si to Orderings( plan)
Promotion: Add Sj < Sthreat to Orderings( plan)
if not Consistent( plan) then fail
end
 Clobbering and
promotion/demotion
• A clobberer is a potentially intervening step that destroys the condition
achieved by a causal link. E.g., Go(Home) clobbers At(Supemarket):

Demotion: put before Go(Supermarket)

Promotion: put after Buy(Milk)

29
Properties of POP
• Nondeterministic algorithm: backtracks at choice points on failure
• choice of Sadd to achieve Sneed
• choice of demotion or promotion for clobberer
• selection of Sneed is irrevocable
• POP is sound, complete, and systematic (no repetition)
• Extensions for disjunction, universals, negation, conditionals
• Can be made effcient with good heuristics derived from problem
description
• Particularly good for problems with many loosely related subgoals

30
Example: Blocks world

31
Example: Blocks world

32
Example: Blocks world

33
Example: Blocks world

34
Example: Blocks world

35
Planning Graphs
• A planning graph consists of a sequence of levels that
correspond to time-steps in the plan
• Level 0 is the initial state.
• Each level contains a set of literals and a set of actions
• Literals are those that could be true at the time step.
• Actions are those that their preconditions could be satisfied
at the time step.
• Works only for propositional planning.

36
Example:Have cake and eat it too

37
 The Planning graphs for “have
cake”,
• Persistence actions: Represent “inactions” by boxes: frame axiom
• Mutual exclusions (mutex) are represented between literals and actions.
• S1 represents multiple states
• Continue until two levels are identical. The graph levels off.
• The graph records the impossibility of certain choices using mutex links.
• Complexity of graph generation: polynomial in number of literals.

38
Defining Mutex relations
• A mutex relation holds between two actions on the same
level iff any of the following holds:
• Inconsistency effect:one action negates the effect of another.
Example “eat cake and presistence of have cake”
• Interference: One of the effect of one action is the negation of the
precondition of the other. Example: eat cake and persistence of Have
cake
• Competing needs: one of the preconditions of one action is
mutually exclusive with a precondition of another. Example:
Bake(cake) and Eat(Cake).
• A mutex relation holds between 2 literals at the same level iff one is
the negation of the other or if each possible pair of actions that can
achieve the 2 literals is mutually exclusive.

39
 Planning graphs for heuristic
estimation
• Estimate the cost of achieving a goal by the level in the
planning graph where it appears.
• To estimate the cost of a conjunction of goals use one of the
following:
• Max-level: take the maximum level of any goal (admissible)
• Sum-cost: Take the sum of levels (inadmissible)
• Set-level: find the level where they all appear without Mutex
• Graph plans are relaxation of the problem.
• Representing more than pair-wise mutex is not cost-effective

40
The graphplan algorithm

41
Planning graph for spare tire a S2
goal: at(spare,axle)
• S2 has all goals and no mutex so we can try to extract solutions
• Use either CSP algorithm with actions as variables
• Or search backwards
Search planning-graph backwards
with heuristics
• How to choose an action during backwards search:
• Use greedy algorithm based on the level cost of the literals.
• For any set of goals:
• 1. Pick first the literal with the highest level cost.
• 2. To achieve the literal, choose the action with the easiest
preconditions first (based on sum or max level of precond
literals).

43
Properties of planning graphs;
termination
• Literals increase monotonically
• Once a literal is in a level it will persist to the next level
• Actions increase monotonically
• Since the precondition of an action was satisfied at a level and
literals persist the action’s precond will be satisfied from now on
• Mutexes decrease monotonically:
• If two actions are mutex at level Si, they will be mutex at all
previous levels at which they both appear
• Because literals increase and mutex decrease it is guaranteed
that we will have a level where all goals are non-mutex

44
Planning with propositional logic
• Express propositional planning as a set of propositions.
• Index propositions with time steps:
• On(A,B)_0, ON(B,C)_0
• Goal conditions: the goal conjuncts at time T, T is
determined arbitrarily.
• Unknown propositions are not stated.
• Propositions known not to be true are stated negatively.
• Actions: a proposition for each action for each time slot.
• Succesor state axioms need to be expressed for each action
(like in the situation calculus but it is propositional)

45
Planning with propositional logic
(continued)
• We write the formula:
• Initial state and successor state axioms and goal
• We search for a model to the formula. Those actions that are
assigned true constitute a plan.
• To have a single plan we may have a mutual exclusion for all
actions in the same time slot.
• We can also choose to allow partial order plans and only
write exclusions between actions that interfere with each
other.
• Planning: iteratively try to find longer and longer plans.

46
SATplan algorithm

47
Complexity of satplan
• The total number of action symbols is:
• |T|x|Act|x|O|^p
• O = number of objects, p is scope of atoms.
• Number of clauses is higher.
• Example: 10 time steps, 12 planes, 30 airports, the complete
action exclusion axiom has 583 million clauses.

48