Downloaded from https://www.everyspec.com AMCP 706-197 I)From these examples we can se that the X -1 optimum of a nonlinear programming prob- lem will not necessuuily be at a comner point of the constraint set and may not even be on the boundary. In addition, there may be local optima distinct from the global optimum. Theme properties are direct consequences of the nonlinearity. However, a damss of nonlin- ear probl-Ims can be defined which ame guaran-b X14 teed to be tree of distinct local optima. These are called convex programming problems.
Before moms of the specific methods of solv- ing constrained minimization problems are described, the concept of convexity and its implications for nonlinear programming will be discussed. Here the constraint set has curved boundaries which cause the local minimum f(x) to be 40; the global 12-3.2 CONVEXITY minimum fAx) in this case is 15. There are several reasons why the con- FIGURE 17-7. Local Minima Due to Curved cepts of convexity and convex functions Constraints' (which will be defined in this paagrah) are important in nonlinear programming. It is lem is convex.
Even though there are many usually impossible to prove that a given proce- ra-ol rbesta r o ovx e dure will find the global minimum of a non- rel-worlid droblmsethty auoticons liner pogramin prblemunlss te pob- often can give insight into the properties of more general problems. Sometimes, such re- sults even can be carried over to problems that are not convex, but in a weaker form. In fact, few important mathematical results have 6- been derived in the programming field with- out assuming convexity.
5.X, X.1 .0 Convexity thus plays a role:~ mathemati- cal programming which is similar to the role of linearty in the study of dynamic systems, 4 frx) 0where many results derived from linear theory are used in the design of nonlinear control systems. The main theorem of convex program- ming is that any local minimum of a convex programming problem is a global minimum. If the protlem has a number of points at which *2.5 X 2 0 the global minimum exists, the set of all such points :S convex, and no distinct, separate, local minima with different functional values NO71h0/2 /X3 4 5 6 can exist.
This is a very convenient property X, since it greatly simplifies the task of locating Thare may be more than one minimum point within the global minimum. the constraint set. Here, fix) = 4 and f~x) = 3 are A set of points is convex if the line seg- both constrained minima, but *~) = 4 is only local. ment joining any two of these points remains Lo~a MinMUM4in the set. In Fig. 12-8, sets A and B are con- Ilk FIGURE 72.6. Loa iiu 4 vex, while C is not. A convex set can be 12-9