Speaker
Description
In high-precision metrology and geodetic network design, the achievable accuracy depends not only on the instruments
but also on the measurement geometry — in particular, on where component fiducials and instrument stations are placed.
These positions are traditionally chosen from experience and have to be fixed before any measurement is taken;
selecting them systematically therefore offers a valuable lever for reducing the measurement uncertainty.
This contribution presents a framework, built on CERN's open-source geodetic least-squares adjustment platform LGC2 (Logiciel Général de Compensation),
that applies Optimum Experimental Design (OED) to this problem.
The estimation variables are split into optimisation variables,
which define the objective function — D-optimality (error-ellipsoid volume) or A-optimality (average parameter variance) —
and bounded design variables that the optimiser searches over.
The framework draws on LGC2's modelling capabilities, particularly its local coordinate systems, called frames, parametrized by estimable parameters.
We illustrate it on a girder fiducialisation scenario, optimising the placement of fiducials on the magnet components and the positioning of the stations
to determine the components' relative position and orientation with minimised uncertainty.
The result is a systematic way to reduce the expected alignment uncertainty at the design stage, before any hardware is installed.
