PowerPoint Presentation
Modelling Preferences in PDDL 3.0
Copyright By PowCoder代写 加微信 powcoder
Goals for today
Get to work
If I do that, get a coffee on the way
If I get a coffee, go to the loo after
Play the piano
Goals for today
Get to work
If I do that, get a coffee on the way
If I get a coffee, go to the loo after
Play the piano
Goals for *the end* of today?
(:goals (and
(at amanda work)
Work/life balance
(sometime (at amanda work))
(sometime (at amanda piano))
Selling out…
(preference p0 (sometime (at amanda work)))
(preference p1 (sometime (at amanda piano)))
cost(p0) = 100
cost(p1) = 5
With goal(s)
(:goal (and
(at amanda mybed)
(:constraints (and
(preference p0 (sometime (at amanda work)))
(preference p1 (sometime (at amanda piano)))
If I do that, get a coffee on the way
(sometime (at amanda coffeeshop)) ?
(sometime-before (at amanda work)
(at amanda coffeeshop))
a = (at amanda work)
b = (at amanda coffeeshop)
If I get a coffee, go to the loo
(sometime (at amanda loo)) ???
If I get a coffee, go to the loo
(sometime-before (at amanda work)
(at amanda coffeeshop))
(sometime-after (at amanda coffeeshop)
(at amanda loo))
a = (at amanda coffeeshop)
b = (at amanda loo)
Sometime-Before vs Sometime-After
(sometime-before (at amanda work)
(at amanda coffeeshop))
(sometime-after (at amanda coffeeshop)
(at amanda work))
These are not inverses:
If I go to work then I have to have coffee first;
If I drink coffee I have to go to work afterwards.
Watch Out!
What if playing the piano adds (played amanda piano):
(sometime-after (at amanda work)
(played amanda piano))
Order a then b then c?
(and (sometime-before (b) (a))
(sometime-after (b) (c))
We can satisfy this with b, c, a, b not just a, b, c!
One last sort of preference
(:goal (and
(at amanda mybed)
(preference p4 (switched-off phone))
Solving Planning Problems With Soft Goals
Goal preferences must be easy?
It’s almost a goal, c’mon…
Compiling goal preferences
RPG Heuristic?
What’s a relaxed plan for a problem with four compiled-away goal preferences?
RPG Heuristic?
What’s a relaxed plan for a problem with four compiled-away goal preferences?
(forgo p0) (forgo p1) (forgo p2) (forgo p3)
h(S) = 5 (‘distance to go’)
Regardless of how close it is to meeting a goal
Keyder & Geffner, JAIR, 2009
程序代写 CS代考 加微信: powcoder QQ: 1823890830 Email: powcoder@163.com