Abstract:

A digital compensation network is designed which (1) minimizes deviation from the user-determined optimal transducer response, (2) limits the sensitivity to physical perturbations of that response, and (3) ensures actuator authority limitations for known and unknown-but-bounded input sequences. H[sup (infinity)], H[sup 2], l[sup (infinity)], and l[sup 1] transfer function norms translate the design goals and specifications into a convex constrained optimization problem, which is solved using semidefinite programming. The methodology can be simply extended to treat transducer arrays and optimal beamforming. A mathematically modeled underwater acoustic transducer provides a numerical example for exploring the optimal trade-off between competing specifications. [This work is partialy supported by SRI International.]