Plan Generation & Causal-Link Planning 1

Plan Generation &
Causal-Link Planning 1
Review
 Frame problem:
 How to specify what does not change
 Qualification problem
 Hard to specify all preconditions
 Ramification problem
 Hard to specify all effects
Review
 Situation Calculus
 Successor state axioms:
Broken(x, do(s,a)) ↔ [a = drop(x) ∧ fragile(x, s)]∨
∃b[a=explode(b) ∧ nextTo(b, x, s)] ∨
broken(x,s) ∧ ¬ ∃ a = repair(x)
 Preconditions axioms:
Poss(pickup(r,x), s) ↔ robot(r) ∧ ∀z ¬holding(z, x, s) ∧
nextTo(r, x, s)
 Strips representation
 Means-ends analysis
 Networks of Actions (Noah)
Domain-Independent
Planning
Inputs:
 Domain Action Theory
 Problem Instance
 Description of (initial) state of the world

 Specification of desired goal behavior
Output: sequence of actions that
executed in initial state satisfy goal
“Classical Planning”
Assumptions
 Atomic Time
 Instantaneous actions
 Deterministic Effects
 Omniscience
 Sole agent of change
 Goals of attainment
Example Problem Instance:
“Sussman Anomaly”
Initial State: Goal:
A
C B
A B C
Initial State: (on-table A) (on C A) (on-table B)

(clear B) (clear C)
Goal: (on A B) (on B C)
Example Problem Instance:
“Sussman Anomaly”
Initial State: Goal:
A
C B
A B C
Initial State: (and (on-table A) (on C A)

(on-table B) (clear B) (clear C))
Goal: (and (on A B) (on B C))
Action Representation:
Propositional STRIPS
Move-C-from-A-to-Table:
preconditions: (on C A) (clear C)
effects:
add (on-table C)
delete (on C A)
add (clear A)
Solution to frame problem: explicit effects are

the only changes to the state.
Action Representation:
Propositional STRIPS
Move-C-from-A-to-Table:
preconditions: (and (on C A) (clear C))
effects:
(and (on-table C)
(not (on C A))
(clear A))
Solution to frame problem: explicit effects are

the only changes to the state.
Plan Generation:
Search space of world states
Planning as a (graph) search problem
 Nodes: world states
 Arcs: actions
 Solution: path from the initial state to

one state that satisfies the goal
 Initial state is fully specified
 There are many goal states
Search Space: Blocks World
Progression:
Forward search (I -> G)
ProgWS(state, goals, actions, path)
If state satisfies goals, then return path
else a = choose (actions), s.t.
preconditions(a) satisfied in state
if no such a, then return failure
else return
ProgWS(apply(a, state), goals, actions,
concatenate(path, a))
First call: ProgWS(IS, G, Actions, ())
Progression Example:
Sussman Anomaly
I: (on-table A) (on C A) (on-table B) (clear B) (clear C)
G: (on A B) (on B C) Non-Deterministic
 P(I, G, BlocksWorldActions, ()) Choice!
 P(S1, G, BWA, (move-C-from-A-to-table))
 P(S2, G, BWA, (move-C-from-A-to-table,
move-B-from-table-to-C))
move-B-from-table-to-C,
G ⊆ S3 => Success!
move-A-from-table-to-B))
Regression:
Backward Search (I <- G)
RegWS(init-state, current-goals, actions, path)
If init-state satisfies current-goals, then return path
else a =choose (actions), s.t. some effect of a
satisfies one of current-goals
If no such a, then return failure [unachievable*]
If some effect of a contradicts some of current-goals,
then return failure [inconsistent state]
CG’ = current-goals – effects(a) + preconditions(a)
If current-goals ⊂ CG’, then return failure [useless*]
RegWS(init-state, CG’, actions, concatenate(a,path))
First call: RegWS(IS, G, Actions, ())
Regression Example:
Sussman Anomaly
 R(I, G, BlocksWorldActions, ()) Choice!
 R(I, ((clear A) (on-table A) (clear B) (on B C)), BWA,
(move-A-from-table-to-B))
 R(I, ((clear A) (on-table A) (clear B) (clear C), (on-table B)),
BWA, (move-B-from-table-to-C, move-A-from-table-to-B))
 R(I, ((on-table A) (clear B) (clear C) (on-table B) (on C A)),
BWA, (move-C-from-A-to-table, move-B-from-table-to-C,
current-goals ⊆ I => Success!

Regression Example:
Sussman Anomaly
 P(I, G, BlocksWorldActions, ()) Choice!
 P(S1, G, BWA, (move-C-from-A-to-table))
move-B-from-table-to-C))
move-B-from-table-to-C,
G ⊆ S3 => Success!
Progression vs. Regression
 Both algorithms are:
 Sound: the result plan is valid
 Complete: if valid plan exists, they find one
 Non-deterministic choice => search!
 Brute force: DFS, BFS, Iterative Deepening, ..,
 Heuristic: A*, IDA*, …
 Complexity: O(bn) worst-case
b = branching factor, n = |“choose”|
 Regression: often smaller b, focused by goals
 Progression: full state to compute heuristics
Total-Order vs Partial-Order Plans
Plan Generation:
Search space of plans
Partial-Order Planning (POP)
 Nodes are partial plans
 Arcs/Transitions are plan refinements
 Solution is a node (not a path).
Principle of “Least commitment”

 e.g. do not commit to an order of
actions until it is required
Terminology
 Step: a step in the partial plan which is
bound to a specific action
 Ordering: s1<s2 s1 must precede s2
 Open conditions: preconditions of the
steps (including goal step)
 Causal link
 A commitment that the condition at Ac
will be made true by Ap
 Requires Ap to cause Q
 Should have an effect Q
 Causal link
 Unsafe link
Partial Plan Representation
 Plan = (A, O, L), where
 A: set of actions in the plan
 O: temporal orderings between actions (a < b)
 L: causal links linking actions via a literal
 Causal Link: Ap Q Ac
Action Ac (consumer) has precondition Q that
is established in the plan by Ap (producer).
(clear b)
move-a-from-b-to-table move-c-from-d-to-b
Partial plan representation
 P = (A, O, L, OC, UL)
 OC: set bof open conditions
 UL: set of unsafe links where P is
deleted by some Sk
 Flaw: Open condition and unsafe link
Partial Plan Representation
 Example
 Flaw:
 Solution plan: A partial plan with no
remaining flaw. Every open condition
must be satisfied by some action. No
unsafe links should exist (i.e the plan is
consistant)
Algorithm (POP)
1. Let P be an initial plan

2. Flaw selection: choose a flaw f (either open condition or unsafe link)
3. Flw resolution:
1. If f is an open condition
2. choose an action S that achieves f
3. If f is an unsafe link
1. choose promotion or demotio
4. Update P
5. Return NULL if no resolution exist
6. If there is no flaw left, return P
7. else go to 2
Example
 Initial plan
 Intermediate partial plan
 For flaw selection we have several
choices
 Flaw selection (open condition, unsafe
link)
 Flaw resolution (how to select partial
plan?)
 Action selection (backtract point)
 Unsafe link selection (backtract point)
Threats to causal links
Step At threatens link (Ap, Q, Ac) if:
1. At has (not Q) as an effect, and
2. At could come between Ap and Ac, i.e.
O ∪ (Ap < At < Ac ) is consistent
What’s an example of an action that

threatens the link example from the
last slide?
Initial Plan
For uniformity, represent initial state and goal
with two special actions:
 A0:
 no preconditions,
 initial state as effects,
 must be the first step in the plan.
 A∞:
 no effects
 goals as preconditions
 must be the last step in the plan.
POP algorithm
POP((A, O, L), agenda, actions)
If agenda = () then return (A, O, L) Termination
Pick (Q, aneed) from agenda Goal Selection
aadd = choose(actions) s.t. Q ∈effects(aadd) Action Selection
If no such action aadd exists, fail.
L’ := L ∪ (aadd, Q, aneed) ; O’ := O ∪ (aadd < aneed)
agenda’ := agenda - (Q, aneed)
If aadd is new, then A := A ∪ aadd and Update goals
∀P ∈preconditions(aadd), add (P, aadd) to agenda’
For every action at that threatens any causal link (ap, Q, ac) in L’
choose to add at < ap or ac < at to O. Protect causal links
If neither choice is consistent, fail.
- Demotion: at < ap
POP((A’, O’, L’), agenda, actions)
- Promotion: ac < at
POP
POP is sound and complete
 POP Plan is a solution if:
 All preconditions are supported (by causal
links), i.e., no open conditions.
 No threats
 Consistent temporal ordering
 By construction, the POP algorithm
reaches a solution plan
POP example:
Sussman Anomaly
A0
(on C A) (on-table A) (on-table B) (clear C) (clear B)
(on A B) (on B C)
Ainf
Work on open precondition
(on B C) and (clear B)
A0
(clear B) (clear C) (on-table B)

A1: move B from Table to C
-(on-table B) -(clear C) (on B C)
(on A B) (on B C)
Ainf
(on A B)
A0

(clear A) (clear B) (on-table A) A1: move B from Table to C
A2: move A from Table to B -(on-table B) -(clear C) (on B C)
-(on-table A) -(clear B) (on A B)
(on A B) (on B C)
Ainf
Protect causal links
A0
(A0, (clear B), A1) theatened by A2

--> Promotion
(on A B) (on B C)
Ainf
(clear A) and protect links
A0
(on C A) (clear C)
A3: move C from A to Table
-(on C A) (on-table C) (clear A)
(on A B) (on B C)
Ainf
Final plan
A0
(on C A) (clear C)
A3: move C from A to Table
-(on C A) (on-table C) (clear A)
(on A B) (on B C)
Ainf
Plartial-Order Planning vs
State-Space Planning
Complexity: O(bn) worst-case
 Non-deterministic choices (n):
 ProgWS, RegWS: n = |actions|
 POP: n = |preconditions| + |link protection|
 Generally an action has several preconditions
 Branching factor (b)
POP has smaller b:
 No backtrack due to goal ordering
 Least commitment: no premature step ordering
 Does POP make the least possible amount of
commitment?

Plan Generation & Causal-Link Planning 1

Uploaded by

Copyright:

Available Formats

Plan Generation & Causal-Link Planning 1

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Plan Generation & Causal-Link Planning 1

Uploaded by

Copyright:

Available Formats

Plan Generation &

 Description of (initial) state of the world

Initial State: (on-table A) (on C A) (on-table B)

Initial State: (and (on-table A) (on C A)

Solution to frame problem: explicit effects are

Solution to frame problem: explicit effects are

 Solution: path from the initial state to

 P(S2, G, BWA, (move-C-from-A-to-table,

current-goals ⊆ I => Success!

 P(S2, G, BWA, (move-C-from-A-to-table,

 Arcs/Transitions are plan refinements

 Solution is a node (not a path).

Principle of “Least commitment”

1. Let P be an initial plan

What’s an example of an action that

(clear B) (clear C) (on-table B)

(clear B) (clear C) (on-table B)

(A0, (clear B), A1) theatened by A2

You might also like