You are on page 1of 23   An Emergency Service Facility Location Problem With Fuzzy Objective and Constraint.

Tatsuo Matsutomi Hiroaki Ishii  Outline 1. Introduction.

2. Formulation of the problem.

3. Fuzzy Decisions.

4. Properties of the chance constraint.

5. Boundary of Feasible Region.

6. Solution Procedure.

7. Conclusion.  Introduction because their locations are known only after a call is been made, they assumed the locations of demand points are reandom variables. And thats why they introduced the fuzzy sets concepts into the problem. they have developed a solution procedure for an ambulance depot location problem with fuzzy a chance constraint and fuzzy objective and constraint.  Formulation Understanding the model 1.

2.

3.

An ambulance must respond to a call for service from the location of the incident. Transport the patient to the nearest hospital. Then it goes back to its depot, and wait for the next call. From above, there are three important times to consider or to minimize:   1. Response time.

2. Time the ambulance takes from the incident location to the nearest hospital.

3. Time for the ambulance to go back to its depot.  Formulation

Notations 1. [Xi,Yi],i=1,2,…,n locations of demand (calls) points. Assumed to be independently, identically distributed (i.i.d).

2. (x,y) locations of a new facility. (emergency depot).

3. (a,b) locations of existing facilities (hospitals).

4. ro maximum time or distance allowed between a demand point and both the depot and the hospital.

5. rc maximum time or distance allowed between the existing facility and the depot.

6. rp maximum time or distance allowed between a demand point and the depot.

7. Di(x,y) distance function from the depot to the demand point plus the distance between the demand point to the hospital.

8. d(x,y) distance function between depot and the hospital.

9. di(x,y) distance function between depot and the demand point.

10. α is a probability level to which the chance constraint should hold.  Formulation The model The problem P was formulated as follows:

P

:

E(

n

i =

Subject to

 Di(x, y)) p ro 1 : d(x, y) p rc Pr [ max di(x, y) ≤ rp] ≥ α (x, y) ∈ [0, L] * [0, M], 0 ≤ α ≤ 1.

p means that the objective function and the

first constraint can be violated up to a certain

threshold.  Formulation Membership function The functions µo (x,y) and µc (x,y) are introduced in-order to reduce the problem P to ordinary equivalent location problem.

µ

o

(

)

x y =

,

µ

r

(

c x

,

= E (

y

)

n

i

=

Di

1

1

0

( r

ro ) / lo

1

0

− −

d x y

(

(

)

,

1

rc

( x , y )), lo

>

0, lc

>

 if r ≥ ro + lo if ro < r < ro +

if

r ro

lc

if

if ro

r

r

r ro

ro

<

if

lo

ro

lo

+

<

) /

0.

+

lo  formulation µo indicates the degree to which the objective is satisfied and lo is a maximum threshold. In other words it means the decision maker thinks he want the value r to be at most ro but he can accept bigger values of r but decreasing the satisfaction degree to the limit of ro+lo. Properties of the membership functions:

1. Decreasing over the interval [ro,ro+lo].

2. Has the value of 0 at ro+lo and 1 at ro. μ
o
1
ro
0
r
lo  Fuzzy Decision Their first step in generating the solution was to reduce the problem P into two problems P1 and P2 as follows:

P1: Maximize { o(x

Pr [max di(x, y)

)

µ

, y

rp]

α

µ

c(x

)}

, y

(x, y) [0, L]*[0, M], 0 α 1

Here they used the minimum-operator.  Fuzzy Decision And For the fuzzy constraint:

 P2 : Maximize {Wo * µ o(x, y) + Wc * µ c(x, y)} S.T : P r [ max di (x, y) ≤ rp] ≥ α (x, y) ∈ [0, L] * [0, M], 0 ≤α ≤ 1  Fuzzy decision In order to solve P1 and P2 they had to transform them into equivalent problem.

 P1 : Maximize λ S.T :
 n E( ∑ Di(x, y)) ≤ ro + lo(1 - λ ) i = 1 d(x , y ) ≤ rc + lc(1 - λ ) Pr [ max di(x, y) ≤ rp] ≥ α (x, y) ∈ [0, L] * [0, M], 0 ≤ α ≤ 1.  Fuzzy decision P2 : Maximize Wo * λo + Wc * λc . S T : n E( ∑ Di(x, y)) ≤ ro + lo(1 - λo) i = 1 d(x, y) ≤ rc + lc(1 - λc) Pr [max di(x, y) ≤ rp] ≥ α (x, y) ∈ [0, L] * [0, M], 0 ≤ λo, λc, α ≤ 1.  Properties of the chance constraint Pr [max

di(x, y) rp] = ( Pr [d( x, y)

rp] ) n

Property 1: The probability Pr[ rp ] is equal to the proportion of the area where a rotated square region (diamond ( x,y )) whose center is (x , y ) is included in the rectangular [ 0,L ] * [ 0,M ] Where Pr[ rp ] = Pr [ d(x,y)]  Properties of the chance constraint A(x,y) is the area of intersection of rectangular [0,L]*[0,M] and the diamond (x,y) rp(x,y)
A(x,y)
L

M

0  Properties of the chance constraint Property 2:The probability Pr[ rp ] takes maximum value where (x,y) = (L/2,M/2) and takes minimum value where (x,y) =(0,0),(L,0),(0,M) and (L,M).  Boundary of Feasible Region From Property 2 we can identify the feasible region where the chance constraint is satisfied. we define the region R1(α,rp) as follows

Rl(

α , rp)

= [ ( x , y)

A(x, y)/L * M

α 1
n
]  Boundary of Feasible Region R1(α,rp) can be identified by following procedure in which we take into account the Property 2 . We assume that a L/2,

b

M/2

STEP 1:If

A

(L,

M)/(L

* M)

α 1
n

then R1(α,rp) coincide with a rectangular area [0,L]*[0,M]. Stop. Otherwise, go to STEP 2 .  Boundary of Feasible Region STEP 2:If

A (L, M/2)/(L 1
* M)
α n

find points p1 = (L,y1) and q1=(xi,M) such that

A (L, y1)/(L

* M)

= 1
α n

, A (x1, M)/(L

* M)

Then find the relation between x and y that

satisfy

to STEP 3. 1

α n

A (x, y) = L * M *

Stop. Otherwise go

,

= 1
α n  Boundary of Feasible Region STEP 3: 1. Find a point q2=(x2,M/2) such that A(x2,M/2)=L*M*(α^1/n). If rp  Solution Method Let ( x * , y * ) denotes an optimal solution of P1 . So we obtain the following procedure: 1. STEP 1: Identify the boundary of R1(α,rp). 2. STEP 2: Find (xo(λ),yo(λ) ).If (xo(λ),yo(λ))ε R1(α,rp) then (x*,y*)=(xo(λ) ,yo(λ) ) , stop. Otherwise. 3. STEP 3: Find a point of contact (xd(λ),yd(λ)) of the boundary of R1(α,rp) and the diamond. If (xd(λ),yd(λ)) satisfies constraint (10) , then (x*,y*) = (xd(λ),yd(λ)), stop. Otherwise, find a point which satisfies constraint (10) and (11) on the boundary of R1(α,rp). Terminate.  Solution Method Problem P2 can be solved almost using the same procedure as in Problem P1 From (13) and (14) , if ellipse (λo) expressed by equation

(x - L/2)

2

/L + (y

M/2)

2

/M

(ro

+

lo(1 - λo))/n - constant

and diamond (λc) with center (a,b) and radius rc+lc*(1-λc) have a point of contact, the value of objective function is maximized.  Solution Method So we find the relation between λo and λc for ellipse (λo) and diamond (λc) having a contact point. Let (x”,y”) denotes a contact points. If (x”,y”)ε R1(α,rp), then input the relation of λo and λc into the objective function of P2, we can easily find an optimal solution.  Conclusion In this paper they developed a solution procedure of an emergency facility location problem with fuzzy objective and constraint. One uses a minimum-operator and the other uses an add-operator.