We solve the chance constrained optimization with convex feasible set through approximating the chance constraint by another convex smooth function. The approximation is based on the numerical properties of the Bernstein polynomial that is capable of effectively controlling the approximation error for both function value and gradient. Thus, we adopt a first-order algorithm to reach a satisfactory solution which is expected to be optimal. When the explicit expression of joint distribution is not available, we then use Monte Carlo approach to numerically evaluate the chance constraint to obtain an optimal solution by probability. Numerical results for known problem instances are presented.
Available at: http://works.bepress.com/lijian_chen/8/