Statistics and machine learning literature has strong parallels

Optimisation

Different parts of the combining scheme may be optimised separately or simultaneously, depending on the motivating problem.

8.4.4 Data fusion

Architectures

Figures 8.11 and 8.12 illustrate the two main architectures for a decentralised distributed detection system – the serial and parallel configurations. We assume that there are L sensors. The observations at each sensor are denoted by yi and the decision at each sensor by ui; i D 1; : : : ; L, where

The parallel system has been widely considered and is shown in Figure 8.12. Each sensor receives a local observation yi, i D 1; : : : ; L, and produces a local decision ui which is sent to the fusion centre. At the fusion centre all the local decisions ui, i D 1; : : : ; L, are combined to obtain a global decision u0. The parallel architecture is

y1 y2 yL