Finding two relevant papers

[ImaraSpeek]

So the FF-replan with hindsight paper and some other paper

[arian]

Probabilistic Planning vs Replanning Iain Little and Sylvie Thiebaux, 2007 http://users.cecs.anu.edu.au/~iain/icaps07.pdf Deze beschrijft een aantal 'interesting domains'

[ImaraSpeek]

in a survey of the seventh intenational planning competition hebben ze het over het feit dat ze van PPDDL naar RDDL in 2011 zijn gegaan en vanwege de vele nieuwe domeinen dat veel planners relyen op de UCT monte carlo search tree http://web.engr.oregonstate.edu/~afern/classes/cs533/notes/uct.pdf In 2011 was de winnaar van de MDP track PROST http://www.informatik.uni-freiburg.de/~ki/papers/keller-eyerich-icaps2012.pdf Misschien is dat wat om naar te kijken als tweede paper?

[ImaraSpeek]

Hier is de source code te vinden: http://prost.informatik.uni-freiburg.de/

[arian]

Glutton/Gourmond: http://www.cs.washington.edu/ai/planning/glutton.html

[arian]

Uit Probabilistic Planning vs Replanning:

Definition 1 (Probabilistically Interesting Problem). A prob- abilistic planning problem is considered to be ‘probabilis- tically interesting’ if and only if it has all of the following structural properties:

1. there are multiple goal trajectories;
2. there is at least one pair of distinct goal trajectories, τ and τ' , that share a common sequence of outcomes for the first n − 1 outcomes, and where τn and τ'n are distinct outcomes of the same action; and
3. there exist two distinct goal trajectories τ and τ' and out- comes o ∈ τ and o' ∈ τ' of two distinct actions a = act(o) and a' = act(o') such that executing a strictly decreases the maximum probability of reaching a state where a can be executed.