- Variant of the [[Simplex Method|simplex method]]
- Used if the origin is not part of the initial basis
- Equality constraints
- Negative right-hand sides
- Constraints in ≥ form
- Analogue to the [[Two-Phase Simplex Method|two-phase simplex method]]
## Example
$
\begin{alignat}{3}
\text{max} \quad &&& \mathrlap{z = 3x_1+5x_2}\\
\text{s.t.} \quad &&&& x_1 &\leq 4\\
\quad &&&& 2x_2 &\leq 12\\
\quad &&&& 3x_1+2x_2 &= 18\\
\quad &&&& x_1,x_2 &\geq 0
\end{alignat}
$
Add slack variables to get standard form:
$
\begin{alignat}{3}
\text{max} \quad &&& \mathrlap{z = 3x_1+5x_2}\\
\text{s.t.} \quad &&&& x_1 + s_1 &= 4\\
\quad &&&& 2x_2 + s_2 &= 12\\
\quad &&&& 3x_1+2x_2 &= 18\\
\quad &&&& x_1,x_2,s_1,s_2 &\geq 0
\end{alignat}
$
Need to add an artificial variable to constraint that was initially an equality. The cost of the artificial is $M$, a very large number:
$
\begin{alignat}{3}
\text{max} \quad &&& \mathrlap{z = 3x_1+5x_2-Ma_1}\\
\text{s.t.} \quad &&&& x_1 + s_1 &= 4\\
\quad &&&& 2x_2 + s_2 &= 12\\
\quad &&&& 3x_1+2x_2 + a_1 &= 18\\
\quad &&&& x_1,x_2,s_1,s_2,a_1 &\geq 0
\end{alignat}
$
- If the objective function is to be maximized, $a_1$ must equal 0 since it will drive the function's value down
- If $a_1 \geq 0$ in the optimal solution, the problem is not feasible