Chapter 5 Robust Performance Tailoring with Tuning - SSL - MIT
Chapter 5 Robust Performance Tailoring with Tuning - SSL - MIT
Chapter 5 Robust Performance Tailoring with Tuning - SSL - MIT
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
algorithms begin at an initial guess for the design variables and then try some number<br />
of configurations until an optimal design is achieved or some termination conditions<br />
are met. The approaches differ in the methods used to move from one iteration to<br />
the next. Gradient-based methods use gradient information to guide the search for<br />
points that satisfy the necessary conditions for optimality. However, if the solution<br />
space is not convex there is no guarantee that a global optimum is found. Heuristic<br />
methods do not require gradients and use randomized algorithms to search for good,<br />
but not necessarily optimal, solutions. These methods are most useful when the<br />
solution space is non-convex and/or highly constrained as the search is able to jump<br />
out of local minima and across islands of feasibility. A more detailed discussion of<br />
the gradient-based algorithms and a simple example is given in Appendix A. Both<br />
a gradient-based method, sequential quadratic programming (SQP), and a heuristic<br />
method, simulated annealing (SA), are applied to the PT SCI problem.<br />
2.4.1 Sequential Quadratic Programming<br />
Sequential quadratic programming (SQP) is a quasi-Newton method developed to<br />
solve constrained optimization problems such as that of Equation 2.1. Constraints<br />
are handled by augmenting the objective function to produce the Lagrangian function:<br />
L (�x, λ) =f (�x)+<br />
m�<br />
λigi (�x) (2.39)<br />
The general goal of SQP is to find the stationary point of this Lagrangian using New-<br />
ton’s method. Therefore, the algorithm is also referred to as the Lagrange-Newton<br />
method. For a detailed discussion of SQP the reader is referred to Fletcher [44].<br />
SQP is chosen as the gradient-based optimization algorithm for the SCI develop-<br />
ment model because it can handle constrained problems <strong>with</strong> nonlinear objectives and<br />
constraints, and is already implemented in the MATLAB optimization toolbox [108].<br />
Also, analytical gradients of the cost function are available, as described above, en-<br />
abling a computationally efficient search.<br />
In the MATLAB SQP implementation there are two levels of iterations. At each<br />
55<br />
i=1