Minimax Search of a Network Steve Alpern Department of Mathematics, - - PowerPoint PPT Presentation

minimax search of a network
SMART_READER_LITE
LIVE PREVIEW

Minimax Search of a Network Steve Alpern Department of Mathematics, - - PowerPoint PPT Presentation

Minimax Search of a Network Steve Alpern Department of Mathematics, LSE Search for Immobile Hider on a Network Every edge e of Q has a length L ( e ) and the total length is denoted by L ( Q ) = : The length of a minimal (Chinese Postman) tour


slide-1
SLIDE 1

Minimax Search of a Network

Steve Alpern Department of Mathematics, LSE

slide-2
SLIDE 2

Search for Immobile Hider on a Network

Every edge e of Q has a length L (e) and the total length is denoted by L (Q) = : The length of a minimal (Chinese Postman) tour is denoted :

2 4 3 3 L(Q)=12 Terminal node O=S(0) Searcher starting node

slide-3
SLIDE 3

Bounds on V=V(Q,O) for a General Network

Theorem (Gal): For any network (Q; O) ; the value V of the search game for an immobile hider satis…es

  • 2 V
  • 2:

The lower bound holds i¤ Q is Eulerian (has Eulerian Tour). The upper bound holds for trees and i¤ Q is Weakly Eulerian (Gal), that is, consists of a disjoint family of Eulerian networks connected in a tree like fashion.

slide-4
SLIDE 4

Equal Branch Density (EBD) Hider Distribution on Trees

The optimal Hider distribution on a tree is the EBD distribution e: At every branch node it assigns probabilities to the branches proportional to their lengths.

O 2 3 1 2 1 1 5, 1/2 5, 1/2 1/6 2/6 1/4 1/4 1/4 1/4 1/6 2/6

slide-5
SLIDE 5

Arc-Adding Lemma: Get Q0 from a Q by adding edge e of length l 0 between points x; y 2 Q: Then

  • 1. V

Q0 V (Q) + 2l; so V Q0 V (Q) if we identify v1; v2 (l = 0)

  • 2. If l dQ (v1; v2) ; then V

Q0 V (Q) : Any hiding strategy on Q does as

well on Q0:

slide-6
SLIDE 6

Weakly Eulerian Networks

De…nition: A network is weakly Eulerian if it contains a set of disjoint Eulerian networks such that shrinking each to a point transforms the network into a tree.

slide-7
SLIDE 7

Proposition (Gal): If Q is a weakly Eulerian network then V = =2:

Q Q* Q**

All three networks have same : V (Q) V (Q) and V (Q) V (Q) Arc- Adding Lemma. V (Q) = /2) (tree).

  • =2 V (Q)

=2:

slide-8
SLIDE 8

Gal’s Theorem

Theorem [3.26]: For any network Q; V = =2 i¤ Q is weakly Eulerian.

slide-9
SLIDE 9

The ‘Three Arc’ Network A O 1 1 1 x H

0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.5 1.0

x f(x)

Best to hide near A. Pick x (on random arc) with probability density f (x) = ex 0 < x < ln 2 :693: Searcher goes to A; back a bit on another arc, back to A; back to O; back towards A. (S. Gal, L. Pavlovic). V = (4 + ln 2) =3 1: 56 < =2:

slide-10
SLIDE 10

3 1 1 1 4 2 a b c

Tree with asymmetric distances (travel times):

  • ut (left) back (right)

(Alpern-Lidbetter Formula): V =

  • 2 + 1

2

X

leaves j

e (j) (d (0; j) d (j; 0)) : (1)