Task 3

This workpackage is devoted to the design of efficient algorithms for optimization of the criteria C(R). We rely on the output of Task 2 to be able to indirectly find good designs R^* for C(R) by solving optimization problems defined in terms of I(R). This is indicated in the Work-packages diagram by the dimmed gray arrow linking Task 1 to Task 3.

Our ultimate goal is to find efficient sequential constrained designs. We approach this problem in two steps:

1. In a first step we consider the realistic situation where constraints are imposed to the feasible designs R (a typical example is the definition of spatial observation strategies for mobile sensors, where observations are periodically taken along a one-dimensional curve in space).
2. In a second step we will explicitly consider sequential designs, whose performance – when constructed optimally – is known to establish an upper bound to fixed designs (they lead in general to better performance), and are thus particularly interesting when efficiency is a major concern. In this case the design R is incrementally defined, as new observations are acquired:

^… R_n*  f(R_n*)  R_n+1*  f(R_n*) ^…