Линейное программирование. Ч.1. Учебное пособие для бакалавриата
.pdf
3. Geometric Method
Remark. The level line is perpendicular to the gradient. Thus, the gradient is the normal vector of the level line.
Remark. The value of objective function increases if the level line moves in the positive direction of the gradient vector.
Indeed, using the properties of the inner product we can rewrite the objective function f in the form
f =n X +n0 .
Or
f = n
X cosα+n0 .
Taking into account that the level line is perpendicular to the gradient, we have
cosα =1 .
So, if
X →max
Then
f →max .
Using this statement, we can formulate an algorithm for a geometric method for solving the linear programming problem of finding the maximum of a linear function.
Algorithm of Geometrical method (maximum)
1)Build the feasible region.
2)If the feasible region is not empty then set aside the gradient vector from the origin.
3)Construct a straight line (the level line) perpendicular to the gradient and move this line in the positive direction of the gradient vector.
4)The last intersection of the level line with the feasible region is the point of maximum.
—21 —
3. Geometric Method
Fig. 3
Similarly, in the case of a linear programming problem for a minimum, we have.
Algorithm of Geometrical method (minimum)
1)Build the feasible region.
2)If the feasible region is not empty then set aside the gradient vector from the origin.
—22 —
3. Geometric Method
3)Construct a straight line (the level line) perpendicular to the gradient and move this line in the positive direction of the gradient vector.
4)The first level line touching of the feasible region is the point of minimum.
Example. Solve the linear programming problem.
f =3x1 +x2 −10 →max(min)
7x1 +x2 ≤29,3x1 +2x2 ≥25,
4x1 −x2 ≤15,x1 ≥0, x2 ≥0.
Solution.
First let us find the feasible region. The boundaries are given by the following equations.
l1 :7x1 +x2 =29 l2 :3x1 +2x2 =25 l3 :4x1 −x2 =15
The vector n =(3;1)T is the gradient vector (the normal vector of the level line).
The first level line touching of the feasible region is the point D (see the Figure 3). So D is the point of minimum. The point D is the intersection of the lines l2 and l3 . To find D, it is necessary to solve the system of linear equations.
3x1 +2x2 =25, .
4x1 −x2 =15.
We obtain
x1 =4, D(4;1)
x2 =1.
— 23 —
3. Geometric Method
Fig. 4
The last level line touching of the feasible region is the point B (see the Figure 3). So, B is the point of maximum. The point B is the intersection of the lines l1 and l3 (B = l1 ∩l3 ). To find D, it is necessary to solve the system of linear equations.
7x1 +x2 =29,
4x1 −x2 =15.
So
x1 =4, B(5;5)
x2 =1.
Now substituting the coordinates of points D and B into the objective function, we find the maximum and minimum.
fmin = f(D) =3x1 +x2 −10 =3 4 +1−10 =3 ,
fmax = f(B) =3x1 +x2 −10 =3 5+5−10 =10 .
Example. Solve the linear programming problem.
— 24 —
3. Geometric Method
f = −x1 +x2 +4 →min (max)
x1 −x2 +x3 =3,−2x1 +x2 +x4 =2,
x1 ≥0, x2 ≥0, x3 ≥0, x4 ≥0.
Solution. The problem is written in the canonicals form. To apply the geometric method, we need to transform the problem into a standard form. We have from the constraints
x3 =3−x1 +x2 ≥0,x4 =2+2x1 −x2 ≥0,
By removing the basic variables x3, x4 from the system of main constraints, we get the linear programming problem written in standard form
f = −x1 +x2 +4 →min (max)
x1 −x2 ≤3,2x1 −x2 ≥ −2,x1 ≥0,x2 ≥0.
Let us find the feasible region. Its boundaries are given by the following equation
l1 :2x1 −x2 = −2, l2 : x1 −x2 =3, l3 : x1 =0,
l4 : x2 =0.
The intersection of these lines belonging to the feasible region gives us the following corner points O(0;0); A(0;2)
; B(3;0).
The gradient vector n =(−1;1)T (see the Figure 5) is perpendicular to the ray BD (n BD). So, the first intersection of the level line with the feasible region is the
— 25 —
3. Geometric Method
ray BD. Thus, all the points of the ray BD are the points of minimum.
Fig.5
Fig. 5
The vector m =(1;1)T is the direction vector of the ray
BD. So
Xmin = B +t m =(3,0)+t(1,1) =(3+t,t), t ≥0.
Or |
=3+t |
x1 |
|
x |
=t, t ≥0 . |
2 |
|
Note that the initial problem has 4 variables. Let us find the basis variables:
x3 =3−x1 +x2 =3−(3+t) +t =0,x4 =2+2x1 −x2 =2+2(3+t) −t =8.
Thus, the point of minimum is
x1 =3+t,x2 =t,x3 =0,
x4 =8 +t, t ≥0.
— 26 —
3. Geometric Method
Let us find the optimal values:
fmin = f (B)= −x1 +x2 +4 = −(3+t)+t +4 =1 .
fmax = +∞,
because the objective function is unbounded on the feasible set.
Questions for self-control
1)What conditions must be fulfilled for the application of the geometric method.?
2)How to construct a region of a linear programming problem in the case of two variables?
3) What is the gradient of a linear function
f=n1x1 +n2x2 ?
4)What equation defines the level line of a linear
function f =n1x1 +n2x2 ?
5)How are the gradient and the level line of a linear function related?
6)How to find the minimum in a linear programming problem using the geometric method?
7)How to find the maximum in a linear programming problem using the geometric method?
8)How to determine with the help of a geometric method that there are infinitely many solutions of a linear programming problem at maximum?
9)How to determine with the help of a geometric method that there are infinitely many solutions of a linear programming problem at minimum?
10)How to determine with the help of a geometric method that the linear programming problem at maximum is unbounded?
—27 —
3. Geometric Method
Exercises for independent work
Solve linear programming problems using the geometric method.
f=2x1 +3x2 →min
x1 −x2 ≤1,
1)4x1 +3x2 ≥18,2x1 +5x2 ≤25,
x1 ≥0, x2 ≥0.
f= x1 +x2 →maxx1 −2x2 ≤2,
2)x1 −x2 ≥ −2,3x1 +x2 ≤18,x1 ≥0, x2 ≥ 0.
f= x1 +3x2 →min
x1 −x2 ≤ −1,
3)3x1 −x2 ≥ −3,2x1 +x2 ≥4,x1 ≥0, x2 ≥0.
f=2x1 +3x2 →max5x1 +3x2 ≤30,
4)x1 −3x2 ≥4,2x1 +3x2 ≥ −9,x1 ≥0, x2 ≥0.
f= −2x1 −x2 →min
x1 −x2 ≤ −1,
5)x1 +x2 ≥2,x1 ≤2,
x1 ≥0, x2 ≥0.
— 28 —
3. Geometric Method
f=2x1 −8x2 →min
2x1 −x2 ≤8,
6)x1 −4x2 ≥ −10,x1 +3x2 ≥11,x1 ≥0, x2 ≥0.
f= −x1 −x2 +1 →max
3x1 −2x2 ≤10,
7)x1 +x2 ≥5,x1 −4x2 ≥ −10,x1 ≥0, x2 ≥0.
f= x1 −8x2 →min
2x1 −x2 ≤1,
8)2x1 +3x2 ≥23,
x1 −2x2 ≥1,x1 ≥0, x2 ≥0.
f = −2x1 +2x2 +x4 +x5 −5 →min (max)
3x1 −2x2 −x3 = −6,
9)5x1 +2x2 +x4 =252x1 −3x2 +x5 =6,
x1 ≥0, x2 ≥0, x3 ≥0, x4 ≥0, x5 ≥0.
f = x2 −x4 +1 →min (max)
x1 +2x2 +x3 = −6,
10)2x1 +x2 −x4 =25x1 +2x2 +x5 =5,
x1 ≥0, x2 ≥0, x3 ≥ 0, x4 ≥ 0, x5 ≥ 0.
— 29 —
4. SIMPLEX METHOD
The simplex method was invented by George Dantzig in 1941. This method is used if the following conditions fulfilled.
1) A linear programming problem is written in the canonicals form.
2)The system of main constrains is solved by JordanGauss method so we know basic and free variables.
3)The free terms of main constrains are non-negative.
4)The objective function is expressed in terms of the free variables.
Thus, the linear programming problem has the following form.
f = −γr +1xr +1 −γr +2xr +2 − −γnxn +γ0 →max(min)
x1 +a1,r +1xr +1 +a1,r +2xr +2 +a1nxn = b1, |
|
||
x2 |
+a2,r +1xr +1 +a2,r +2xr +2 +a2nxn = b2, |
(4) |
|
|
|
|
|
................................................................. |
|
|
|
|
+ar,r +1xr +1 +ar,r +2xr +2 +ar,nxn = br , |
|
|
xr |
|
||
x1 ≥0,x2 ≥0, ,xn ≥0. |
|
|
|
|
|
|
|
where |
x1; ;xr are the basic variables, |
xr +1; ;xn |
are the |
free variables, bi ≥0 (i =1; ;r ). |
problem |
(4) has |
|
Remark. The linear programming |
|||
a feasible solution (we will call it the basic solution)
X1 =(b1,b2, ,br ,0, ,0)≥0 .
This solution is obtained from the constraints under the assumption that all free variables are equal to zero.
— 30 —
