Chapter 5 Robust Performance Tailoring with Tuning - SSL - MIT

and the worst-case performance is above that of the RPT design. The tuned perfor-
mance of the RPTT designs, however, is better than that of either the PT or RPT
designs for all uncertainty levels. The RPTT design further extends the performance
of the system at a given uncertainty level, and is applicable to all design regimes de-
fined by combinations of PT, RPT and tuning. Random hardware simulations of the
three designs are run to evaluate the design performance over the entire uncertainty
space. The RPTT design is the clear winner in that it is the only design for which
all hardware simulations either meet the requirements nominally or after hardware
tuning is employed.
7.2 Contributions
This thesis develops a novel approach to the design of high-performance, high-uncertainty
systems in which dynamic tuning is anticipated and formalized in a design optimiza-
tion. This design methodology, RPTT, is especially applicable to high-precision opti-
cal space systems, such as the Space Interferometry Mission (SIM), the James Webb
Space Telescope (JWST) and the Terrestrial Planet Finder (TPF). Specific thesis
contributions are as follows:
• Development of a design methodology, RPTT, that formalizes a complimen-
tary relationship between dynamic tailoring and tuning. RPTT extends robust
design for application to systems that require high levels of performance and
exhibit high uncertainty.
• Development of a model updating technique for application to dynamic tai-
loring that utilizes limited hardware performance data and isoperformance to
reduce the parametric uncertainty space so that a robust tuning optimization
performed on the model yields tuning parameter values that successfully tune
the hardware.
• Study of gradient-based and heuristic optimization techniques for application
to tailoring and tuning given a dynamic performance model. Eigenvector and
215