C1
C1 C2
100
C3
0
C4
30
C5
25
C6
0
C7
0
C8
0
C9
0
∑M ij
p
C2 100 80 40 20 30 0 0 0
C3 0 80 0 10 30 0 0 0
C4 30 40 0 120 0 80 25 0 S' S'
C5 25 20 10 120 15 160 30 50 Where p and q are two different set in a covering of
C6 0 30 30 0 15 0 100 60 the set of the n cells. After the construction of the above
C7 0 0 0 80 160 0 40 0 objective function, we give a heuristic algorithm to get a
C8 0 0 0 25 30 100 40 75
C9 0 0 0 0 50 60 0 75
nearly-ideal covering of the set of n cells ( leaf location
In the following sections, we give a set-covering databases corresponding to the n cells) with the objective to
algorithm to organize the location databases with the get the minimal value of num-crossing.
objective to minimize the total location management time Then each component (a leaf location database set which
cost. is composed of d(explained later) location databases) of the
cover forms a cluster. The following is the procedure to get
Definition 3 A set of sets S1, S2,…, Sn is called a cover of a the non-overlapping cover. In the following section, the
n number of elements in a set S is denoted as |S|. d is the
set S if and only if S = ∪ Si number of sub-location databases in the hierarchical tree. In
i =1 figure1, d is 3.
Definition 4 A set of sets S1, S2,…, Sn is called a PROCEDURE LDB-structuring
non-overlapping cover of a set S if and only if INPUT: Mij, S1, S2, …., Sn
n OUTPUT: A topology structure for the location
(1) S = ∪ Si database tree
i =1
setid=1;
(2) ∀i, j ,1 ≤ i, j ≤ n, i ≠ j , Si ∩ Sj = φ
In general, a (non-overlapping) cover of a set is associated
REPEAT UNTIL ∪S =φ
i
i
40
RC
30
CSC
20
10
0
2 4 6 8 10
The Mean Value of NBC