novel is not always better on the relation between
play

Novel Is Not Always Better: On the Relation between Novelty and - PowerPoint PPT Presentation

Classical Planning Novelty Dominance Relation Novelty Heuristics Conclusions Novel Is Not Always Better: On the Relation between Novelty and Dominance Pruning Joschka Gro, Alvaro Torralba, Maximilian Fickert Classical Planning


  1. Classical Planning Novelty Dominance Relation Novelty Heuristics Conclusions Novel Is Not Always Better: On the Relation between Novelty and Dominance Pruning Joschka Groß, ´ Alvaro Torralba, Maximilian Fickert

  2. Classical Planning Novelty Dominance Relation Novelty Heuristics Conclusions Classical Planning Definition. A planning task is a 4-tuple Π = ( V , A , I , G ) where: • V is a set of state variables, each v ∈ V with a finite domain D v . • A is a set of actions; each a ∈ A is a triple ( pre a , eff a , c a ) , of precondition and effect (partial assignments), and the action’s cost c a ∈ R + 0 . • Initial state I (complete assignment), goal G (partial assignment). → Solution (“Plan”): Action sequence mapping I into s s.t. s | = G . Groß, Torralba, Fickert Novel Is Not Always Better 2/18

  3. Classical Planning Novelty Dominance Relation Novelty Heuristics Conclusions Classical Planning Definition. A planning task is a 4-tuple Π = ( V , A , I , G ) where: • V is a set of state variables, each v ∈ V with a finite domain D v . • A is a set of actions; each a ∈ A is a triple ( pre a , eff a , c a ) , of precondition and effect (partial assignments), and the action’s cost c a ∈ R + 0 . • Initial state I (complete assignment), goal G (partial assignment). → Solution (“Plan”): Action sequence mapping I into s s.t. s | = G . 100 Running Example: A B • V = { t , p 1 , p 2 , f } with D t = { A , B } and D p i = { t , A , B } , D f = { 100 , 99 , 98 , . . . , 0 } . • A = { load ( p i , x ) , unload ( p i , x ) , drive ( x , x ′ ) } Groß, Torralba, Fickert Novel Is Not Always Better 2/18

  4. Classical Planning Novelty Dominance Relation Novelty Heuristics Conclusions What this is about? Novelty (Lipovetzky and Geffner, 2012) (Lipovetzky and Geffner, 2017) (Katz, Lipovetzky, Moshkovich and Tuisov 2017) (Fickert 2018) A (pruning) technique which has greatly improved the state of the art in satisficing planning Dominance (Torralba and Hoffmann, 2015), (Torralba, 2017), (Torralba, 2018): A safe pruning technique for cost-optimal planning Groß, Torralba, Fickert Novel Is Not Always Better 3/18

  5. Classical Planning Novelty Dominance Relation Novelty Heuristics Conclusions Novelty The novelty of s N ( s ) is defined to be the size of the smallest fact set it produces for the first time. Groß, Torralba, Fickert Novel Is Not Always Better 4/18

  6. Classical Planning Novelty Dominance Relation Novelty Heuristics Conclusions Novelty The novelty of s N ( s ) is defined to be the size of the smallest fact set it produces for the first time. IW(K): Breadth first search, pruning all s with N ( s ) > k • Polynomial time • No guidance towards the goal • Good for exploration/achieving single goal facts Groß, Torralba, Fickert Novel Is Not Always Better 4/18

  7. Classical Planning Novelty Dominance Relation Novelty Heuristics Conclusions Novelty The novelty of s N ( s ) is defined to be the size of the smallest fact set it produces for the first time. IW(K): Breadth first search, pruning all s with N ( s ) > k • Polynomial time • No guidance towards the goal • Good for exploration/achieving single goal facts Novelty Heuristics: • Combine the definition of novelty with heuristics • State of the art in satisficing planning Groß, Torralba, Fickert Novel Is Not Always Better 4/18

  8. Classical Planning Novelty Dominance Relation Novelty Heuristics Conclusions Novelty The novelty of s N ( s ) is defined to be the size of the smallest fact set it produces for the first time. IW(K): Breadth first search, pruning all s with N ( s ) > k • Polynomial time • No guidance towards the goal • Good for exploration/achieving single goal facts Novelty Heuristics: • Combine the definition of novelty with heuristics • State of the art in satisficing planning But, why is novelty so good? Groß, Torralba, Fickert Novel Is Not Always Better 4/18

  9. Classical Planning Novelty Dominance Relation Novelty Heuristics Conclusions Example IW(1) 100 A B A B A B T 100 99 98 97 x x x Groß, Torralba, Fickert Novel Is Not Always Better 5/18

  10. Classical Planning Novelty Dominance Relation Novelty Heuristics Conclusions Example IW(1) 100 load ( p 1 ) A B 100 A B A B A B T 100 99 98 97 x x x x Groß, Torralba, Fickert Novel Is Not Always Better 5/18

  11. Classical Planning Novelty Dominance Relation Novelty Heuristics Conclusions Example IW(1) 100 load ( p 1 ) A B 100 A B 99 drive ( A , B ) A B A B A B T 100 99 98 97 x x x x x x Groß, Torralba, Fickert Novel Is Not Always Better 5/18

  12. Classical Planning Novelty Dominance Relation Novelty Heuristics Conclusions Example IW(1) 100 A B unload ( p 1 ) 100 load ( p 1 ) A B 100 A B 99 drive ( A , B ) A B A B A B T 100 99 98 97 x x x x x x Groß, Torralba, Fickert Novel Is Not Always Better 5/18

  13. Classical Planning Novelty Dominance Relation Novelty Heuristics Conclusions Example IW(1) 100 A B unload ( p 1 ) 100 load ( p 1 ) A B 100 A B 99 drive ( A , B ) A B A B A B T 100 99 98 97 x x x x x x Groß, Torralba, Fickert Novel Is Not Always Better 5/18

  14. Classical Planning Novelty Dominance Relation Novelty Heuristics Conclusions Example IW(1) 100 A B unload ( p 1 ) 99 100 load ( p 1 ) A B A B drive ( A , B ) 100 A B 99 drive ( A , B ) A B A B A B T 100 99 98 97 x x x x x x Groß, Torralba, Fickert Novel Is Not Always Better 5/18

  15. Classical Planning Novelty Dominance Relation Novelty Heuristics Conclusions Example IW(1) 100 A B unload ( p 1 ) 99 100 load ( p 1 ) A B A B drive ( A , B ) 100 A B 99 drive ( A , B ) A B A B A B T 100 99 98 97 x x x x x x Groß, Torralba, Fickert Novel Is Not Always Better 5/18

  16. Classical Planning Novelty Dominance Relation Novelty Heuristics Conclusions Example IW(1) 100 A B unload ( p 1 ) 99 100 load ( p 1 ) A B A B drive ( A , B ) 100 A B 99 98 drive ( A , B ) A B A B drive ( B , A ) A B A B T 100 99 98 97 x x x x x x x Groß, Torralba, Fickert Novel Is Not Always Better 5/18

  17. Classical Planning Novelty Dominance Relation Novelty Heuristics Conclusions Example IW(1) 100 A B unload ( p 1 ) 99 100 98 load ( p 1 ) A B A B A B load ( p 1 ) drive ( A , B ) 100 A B 99 98 drive ( A , B ) A B A B drive ( B , A ) A B A B T 100 99 98 97 x x x x x x x Groß, Torralba, Fickert Novel Is Not Always Better 5/18

  18. Classical Planning Novelty Dominance Relation Novelty Heuristics Conclusions Example IW(1) 100 A B unload ( p 1 ) 99 100 98 load ( p 1 ) A B A B A B load ( p 1 ) drive ( A , B ) 100 A B 99 98 drive ( A , B ) A B A B drive ( B , A ) A B A B T 100 99 98 97 x x x x x x x Groß, Torralba, Fickert Novel Is Not Always Better 5/18

  19. Classical Planning Novelty Dominance Relation Novelty Heuristics Conclusions Example IW(1) 100 A B unload ( p 1 ) 99 100 98 load ( p 1 ) A B A B A B load ( p 1 ) drive ( A , B ) 100 A B 99 97 98 drive ( A , B ) A B A B A B drive ( B , A ) drive ( A , B ) A B A B T 100 99 98 97 x x x x x x x x Groß, Torralba, Fickert Novel Is Not Always Better 5/18

  20. Classical Planning Novelty Dominance Relation Novelty Heuristics Conclusions Dominance Analysis Compare states: Which one is better? s t 50 100 A B A B Groß, Torralba, Fickert Novel Is Not Always Better 6/18

  21. Classical Planning Novelty Dominance Relation Novelty Heuristics Conclusions Dominance Analysis Compare states: Which one is better? s t 50 100 A B A B Dominance Relation If s � t , then h ∗ ( s ) ≥ h ∗ ( t ) : t is at least as good as s Groß, Torralba, Fickert Novel Is Not Always Better 6/18

  22. Classical Planning Novelty Dominance Relation Novelty Heuristics Conclusions Dominance Analysis Compare states: Which one is better? s t 50 100 A B A B Dominance Relation If s � t , then h ∗ ( s ) ≥ h ∗ ( t ) : t is at least as good as s Groß, Torralba, Fickert Novel Is Not Always Better 6/18

  23. Classical Planning Novelty Dominance Relation Novelty Heuristics Conclusions Dominance Analysis Compare states: Which one is better? s t 50 100 A B A B Dominance Relation If s � t , then h ∗ ( s ) ≥ h ∗ ( t ) : t is at least as good as s → We can reason about variables independently! Groß, Torralba, Fickert Novel Is Not Always Better 6/18

  24. Classical Planning Novelty Dominance Relation Novelty Heuristics Conclusions Dominance Analysis Compare states: Which one is better? s t 50 100 A B A B Dominance Relation If s � t , then h ∗ ( s ) ≥ h ∗ ( t ) : t is at least as good as s → We can reason about variables independently! : A � B : A � B 0 � 1 � 2 � 3 . . . (no matter the position of other packages or trucks) Groß, Torralba, Fickert Novel Is Not Always Better 6/18

  25. Classical Planning Novelty Dominance Relation Novelty Heuristics Conclusions Dominance Pruning Prune s if there exists t s.t. g ( t ) ≤ g ( s ) and s � t 100 A B Groß, Torralba, Fickert Novel Is Not Always Better 7/18

Recommend


More recommend