properties of heuristics that guarantee a* finds optimal paths robert holte this talk:...
Post on 21-Dec-2015
228 views
TRANSCRIPT
![Page 1: Properties of Heuristics that Guarantee A* Finds Optimal Paths Robert Holte this talk: holte/CMPUT651/admissibility.ppt](https://reader030.vdocuments.mx/reader030/viewer/2022013108/56649d6b5503460f94a4a901/html5/thumbnails/1.jpg)
Properties of Heuristics that Guarantee A* Finds Optimal Paths
Robert Holte
this talk: http://www.cs.ualberta.ca/~holte/CMPUT651/admissibility.ppt
![Page 2: Properties of Heuristics that Guarantee A* Finds Optimal Paths Robert Holte this talk: holte/CMPUT651/admissibility.ppt](https://reader030.vdocuments.mx/reader030/viewer/2022013108/56649d6b5503460f94a4a901/html5/thumbnails/2.jpg)
Best-first Search
• Open list of nodes reached but not yet expanded• Closed list of nodes that have been expanded• Choose lowest cost node on Open list• Add it to Closed, add its successors to Open• Stop when Goal is first removed from Open
Dijkstra: cost, f(N) = g(N) = distance from start
A*: cost, f(N) = g(N) + h(N)
![Page 3: Properties of Heuristics that Guarantee A* Finds Optimal Paths Robert Holte this talk: holte/CMPUT651/admissibility.ppt](https://reader030.vdocuments.mx/reader030/viewer/2022013108/56649d6b5503460f94a4a901/html5/thumbnails/3.jpg)
A* must re-open closed nodes
OPEN: (S,70)CLOSED:
C D GS
A
B
10 10
10
2020
90
h=20
h=110
h=70 h=70 h=60
f = 110
![Page 4: Properties of Heuristics that Guarantee A* Finds Optimal Paths Robert Holte this talk: holte/CMPUT651/admissibility.ppt](https://reader030.vdocuments.mx/reader030/viewer/2022013108/56649d6b5503460f94a4a901/html5/thumbnails/4.jpg)
A* must re-open closed nodes
OPEN: (A,120), (B,40)CLOSED: (S,70)
C D GS
A
B
10 10
10
2020
90
h=20
h=110
h=70 h=70 h=60
f = 110
![Page 5: Properties of Heuristics that Guarantee A* Finds Optimal Paths Robert Holte this talk: holte/CMPUT651/admissibility.ppt](https://reader030.vdocuments.mx/reader030/viewer/2022013108/56649d6b5503460f94a4a901/html5/thumbnails/5.jpg)
A* must re-open closed nodes
OPEN: (A,120), (C,110)CLOSED: (S,70), (B,40)
C D GS
A
B
10 10
10
2020
90
h=20
h=110
h=70 h=70 h=60
f = 110
![Page 6: Properties of Heuristics that Guarantee A* Finds Optimal Paths Robert Holte this talk: holte/CMPUT651/admissibility.ppt](https://reader030.vdocuments.mx/reader030/viewer/2022013108/56649d6b5503460f94a4a901/html5/thumbnails/6.jpg)
A* must re-open closed nodes
OPEN: (A,120), (D,110)CLOSED: (S,70), (B,40), (C,110)
C D GS
A
B
10 10
10
2020
90
h=20
h=110
h=70 h=70 h=60
f = 110
![Page 7: Properties of Heuristics that Guarantee A* Finds Optimal Paths Robert Holte this talk: holte/CMPUT651/admissibility.ppt](https://reader030.vdocuments.mx/reader030/viewer/2022013108/56649d6b5503460f94a4a901/html5/thumbnails/7.jpg)
A* must re-open closed nodes
OPEN: (A,120), (G,140), (subtree with f=110)CLOSED: (S,70), (B,40), (C,110), (D,110)
C D GS
A
B
10 10
10
2020
90
h=20
h=110
h=70 h=70 h=60
f = 110
![Page 8: Properties of Heuristics that Guarantee A* Finds Optimal Paths Robert Holte this talk: holte/CMPUT651/admissibility.ppt](https://reader030.vdocuments.mx/reader030/viewer/2022013108/56649d6b5503460f94a4a901/html5/thumbnails/8.jpg)
A* must re-open closed nodes
OPEN: (A,120), (G,140)CLOSED: (S,70), (B,40), (C,110), (D,110), …
C D GS
A
B
10 10
10
2020
90
h=20
h=110
h=70 h=70 h=60
f = 110
![Page 9: Properties of Heuristics that Guarantee A* Finds Optimal Paths Robert Holte this talk: holte/CMPUT651/admissibility.ppt](https://reader030.vdocuments.mx/reader030/viewer/2022013108/56649d6b5503460f94a4a901/html5/thumbnails/9.jpg)
A* must re-open closed nodes
OPEN: (G,140), (C,90)CLOSED: (S,70), (B,40), (C,110) , (D,110), …(A,120)
C D GS
A
B
10 10
10
2020
90
h=20
h=110
h=70 h=70 h=60
f = 110
![Page 10: Properties of Heuristics that Guarantee A* Finds Optimal Paths Robert Holte this talk: holte/CMPUT651/admissibility.ppt](https://reader030.vdocuments.mx/reader030/viewer/2022013108/56649d6b5503460f94a4a901/html5/thumbnails/10.jpg)
Today’s Question
When a node is first removed from Open, under what conditions are we guaranteed that this path to the node is optimal ?
Dijkstra: all edge-weights are non-negative
A*: the heuristic must have certain properties
![Page 11: Properties of Heuristics that Guarantee A* Finds Optimal Paths Robert Holte this talk: holte/CMPUT651/admissibility.ppt](https://reader030.vdocuments.mx/reader030/viewer/2022013108/56649d6b5503460f94a4a901/html5/thumbnails/11.jpg)
Optimal Path to goal is the first off the Open list
S-N-G optimal, <N, g*(N)+h(N) > is on Open<G,P> on Open is suboptimal
g*(N)+h*(N) < P h*(N) < P – g*(N)
S G
N
P
g*(N) h*(N)
![Page 12: Properties of Heuristics that Guarantee A* Finds Optimal Paths Robert Holte this talk: holte/CMPUT651/admissibility.ppt](https://reader030.vdocuments.mx/reader030/viewer/2022013108/56649d6b5503460f94a4a901/html5/thumbnails/12.jpg)
Admissible Heuristic
Require <N, g*(N)+h(N) > lower cost than <G,P> g*(N)+h(N) < P h(N) < P – g*(N) h(N) h*(N) (because h*(N) < P – g*(N))
A heuristic is admissible if h(N) h*(N) for all N.
Admissible first path to goal off Open is optimal
S G
N
P
g*(N) h*(N)
![Page 13: Properties of Heuristics that Guarantee A* Finds Optimal Paths Robert Holte this talk: holte/CMPUT651/admissibility.ppt](https://reader030.vdocuments.mx/reader030/viewer/2022013108/56649d6b5503460f94a4a901/html5/thumbnails/13.jpg)
Optimal Path to X is the first off the Open list, for all X
S-N-X optimal, <N, g*(N)+h(N) > is on Open<X,P+h(X)> on Open, P is suboptimal
g*(N)+c(N,X) < P c(N,X) < P – g*(N)
S X
N
P
g*(N) c(N,X)
![Page 14: Properties of Heuristics that Guarantee A* Finds Optimal Paths Robert Holte this talk: holte/CMPUT651/admissibility.ppt](https://reader030.vdocuments.mx/reader030/viewer/2022013108/56649d6b5503460f94a4a901/html5/thumbnails/14.jpg)
Consistent Heuristic
Require <N, g*(N)+h(N) > lower cost than <X,P+h(X)> g*(N)+h(N) < P+h(X) h(N) – h(X) < P – g*(N) h(N) – h(X) c(N,X) (because c(N,X) < P – g*(N))
A heuristic is consistent if h(N) c(N,X) + h(X) for all X and all N.
Consistent first path to X off Open is optimal for all X
S X
N
P
g*(N) c(N,X)
![Page 15: Properties of Heuristics that Guarantee A* Finds Optimal Paths Robert Holte this talk: holte/CMPUT651/admissibility.ppt](https://reader030.vdocuments.mx/reader030/viewer/2022013108/56649d6b5503460f94a4a901/html5/thumbnails/15.jpg)
Transforming heuristics into edge weights
Aim: replace the given edge weights and heuristics values with a set of edge weights (and NO heuristic) so that Dijkstra-costs on the new graph are identical to A*-costs on the given graph+heuristic
S A Ba b
A* cost: h(S) a+h(A) a+b+h(B)
![Page 16: Properties of Heuristics that Guarantee A* Finds Optimal Paths Robert Holte this talk: holte/CMPUT651/admissibility.ppt](https://reader030.vdocuments.mx/reader030/viewer/2022013108/56649d6b5503460f94a4a901/html5/thumbnails/16.jpg)
Transformation - goal
S A Ba b
A* cost: h(S) a+h(A) a+b+h(B)
S A B?? ??
Dijkstra cost: h(S) a+h(A) a+b+h(B)
![Page 17: Properties of Heuristics that Guarantee A* Finds Optimal Paths Robert Holte this talk: holte/CMPUT651/admissibility.ppt](https://reader030.vdocuments.mx/reader030/viewer/2022013108/56649d6b5503460f94a4a901/html5/thumbnails/17.jpg)
Transformation (1)
S A Ba b
A* cost: h(S) a+h(A) a+b+h(B)
S A B?? ??
Dijkstra cost: h(S) a+h(A) a+b+h(B)
h(S)
![Page 18: Properties of Heuristics that Guarantee A* Finds Optimal Paths Robert Holte this talk: holte/CMPUT651/admissibility.ppt](https://reader030.vdocuments.mx/reader030/viewer/2022013108/56649d6b5503460f94a4a901/html5/thumbnails/18.jpg)
Transformation (2)
S A Ba b
A* cost: h(S) a+h(A) a+b+h(B)
S A Ba+h(A)-h(S) ??
Dijkstra cost: h(S) a+h(A) a+b+h(B)
h(S)
![Page 19: Properties of Heuristics that Guarantee A* Finds Optimal Paths Robert Holte this talk: holte/CMPUT651/admissibility.ppt](https://reader030.vdocuments.mx/reader030/viewer/2022013108/56649d6b5503460f94a4a901/html5/thumbnails/19.jpg)
Transformation (3)
S A Ba b
A* cost: h(S) a+h(A) a+b+h(B)
S A Ba+h(A)-h(S) b+h(B)-h(A)
Dijkstra cost: h(S) a+h(A) a+b+h(B)
h(S)
![Page 20: Properties of Heuristics that Guarantee A* Finds Optimal Paths Robert Holte this talk: holte/CMPUT651/admissibility.ppt](https://reader030.vdocuments.mx/reader030/viewer/2022013108/56649d6b5503460f94a4a901/html5/thumbnails/20.jpg)
Transformation - general
The order in which nodes come off the Open list using Dijkstra on the transformed graph is identical to the order using A* on the original graph+heuristic.
c(N,X)N X + heuristic
is transformed into
c(N,X) – h(N) + h(X)N X (no heuristic)
![Page 21: Properties of Heuristics that Guarantee A* Finds Optimal Paths Robert Holte this talk: holte/CMPUT651/admissibility.ppt](https://reader030.vdocuments.mx/reader030/viewer/2022013108/56649d6b5503460f94a4a901/html5/thumbnails/21.jpg)
Local Consistency
If edge weights are non-negative, the first path to any node Z that Dijkstra takes off Open is an optimal path to Z.
Non-negative edge weights requires:For all N, and all successors, X, of N
0 c(N,X) – h(N) + h(X) h(N) c(N,X) + h(X)
A heuristic is locally consistent if h(N) c(N,X) + h(X) for all N and all successors X of N.
Locally consistent consistent
c(N,X) – h(N) + h(X)N X
![Page 22: Properties of Heuristics that Guarantee A* Finds Optimal Paths Robert Holte this talk: holte/CMPUT651/admissibility.ppt](https://reader030.vdocuments.mx/reader030/viewer/2022013108/56649d6b5503460f94a4a901/html5/thumbnails/22.jpg)
MonotonicityWith Dijkstra and non-negative edge weights, cost
cannot decrease along a path since it is just the sum of the edge weights along the path.
Because A* with a consistent heuristic is equivalent to Dijkstra with non-negative edge weights, it follows that A*costs along a path can never decrease if the heuristic is consistent.
S A B
A* cost: f(S) f(A) f(B)
![Page 23: Properties of Heuristics that Guarantee A* Finds Optimal Paths Robert Holte this talk: holte/CMPUT651/admissibility.ppt](https://reader030.vdocuments.mx/reader030/viewer/2022013108/56649d6b5503460f94a4a901/html5/thumbnails/23.jpg)
Admissibility Monotonicity
Along path S-A-C, f-values are not monotonic non-decreasing.
/
C D GS
A
B
10 10
10
2020
90
h=20
h=110
h=70 h=70 h=60
f = 110
![Page 24: Properties of Heuristics that Guarantee A* Finds Optimal Paths Robert Holte this talk: holte/CMPUT651/admissibility.ppt](https://reader030.vdocuments.mx/reader030/viewer/2022013108/56649d6b5503460f94a4a901/html5/thumbnails/24.jpg)
Enforced monotonicityCan enforce monotonicity along a path by using
parent’s f-value if it is greater than the child’s f-value.
(valid if h is admissible because the f values on a path never overestimate the path’s true length)
But this does not solve the problem of having to re-open closed nodes in our example.
use f = 8 from parent
f = 8 f = 4
![Page 25: Properties of Heuristics that Guarantee A* Finds Optimal Paths Robert Holte this talk: holte/CMPUT651/admissibility.ppt](https://reader030.vdocuments.mx/reader030/viewer/2022013108/56649d6b5503460f94a4a901/html5/thumbnails/25.jpg)
Summary of definitions
• An admissible heuristic never overestimates distance to goal
• A consistent heuristic obeys a kind of triangle inequality
• With a locally consistent heuristic, h does not decrease faster than g increases
• Monotonicity: costs along a path never decrease
![Page 26: Properties of Heuristics that Guarantee A* Finds Optimal Paths Robert Holte this talk: holte/CMPUT651/admissibility.ppt](https://reader030.vdocuments.mx/reader030/viewer/2022013108/56649d6b5503460f94a4a901/html5/thumbnails/26.jpg)
Summary of Positive Results
• Consistent locally consistent
• Consistent monotonicity
• Consistent admissible
• Consistent first path to X off Open is optimal, for all X
• Admissible first path to Goal off Open is optimal (correctness of the A* stopping condition)
![Page 27: Properties of Heuristics that Guarantee A* Finds Optimal Paths Robert Holte this talk: holte/CMPUT651/admissibility.ppt](https://reader030.vdocuments.mx/reader030/viewer/2022013108/56649d6b5503460f94a4a901/html5/thumbnails/27.jpg)
Summary of Negative Results
• Admissible monotonicity
• Admissible consistent
• Admissible first path to X off Open is optimal, for all X
/
/
/