ABSTRACT
Integer programming involves the use of an algorithm that forces the solution obtained to integer values. Techniques used therein involve the solution of a series of optimization subproblems. The subproblems are obtained by imposing additional constraints on the original problem. The conventional way of obtaining the integer solution by rounding off the continuous solution involves making a decision as to which design variables should be rounded off to the next lower or higher integer. Usually there is no rational way to make this decision. Secondly, the rounded-off solution may be in the infeasible region and moving from the infeasible region to the feasible region could be difficult. More importantly the true integer solution to the nonlinear mixed integer programming problem has a better chance of being globally optimal when the solution space is concave. This follows from the fact that the subproblems solved have disjoint feasible spaces.
