General case

General Case

**Figure 34:** Quantified adaptive fusion (general case with *strong agreement*).
$\begin{figure}\begin{center}\fbox{\epsfbox{c3-adq3.eps}}\end{center}\end{figure}$

**Figure 35:** Quantified adaptive fusion (general case with *weak agreement*).
$\begin{figure}\begin{center}\fbox{\epsfbox{c3-adq4.eps}}\end{center}\end{figure}$

A majority of sources agree at least a little. In figures 34 and 35, the number of agreeing sources is estimated to be at least m=2 sources ( $h(\pi_1, \pi_2) = 1$ ) and at most n=3 sources ( $h(\pi_1, \pi_2, \pi_3)=0.75 > 0$ on the one hand, and $h(\pi_1, \pi_2, \pi_3)=0.35 > 0$ on the other hand).

Area C corresponds to the values considered to be plausible for parameter x by a majority of sources. This area is favoured by using renormalized conjunctive fusion.

Agreeing areas are still found outside area C, but for a lower number of sources. In order not to perform a purely disjunctive fusion, conjunctive fusion is used again, but for a restricted number of sources. Pessimistic estimation m of the number of reliable sources is carried out. Only areas with intersections of m sources are preserved (areas D); and to limit their weight compared to the majority of agreeing sources, their effects are limited to the degree of conflict between sources n.

Thus, if sources n agree strongly ( $h(n) \simeq 1$ ), then sources m, limited to ( $h(n) \simeq 1$ ), are little used (figure 34). However, if sources n, although more numerous, agree little ( $h(n) \simeq 0$ ), then they should be counterbalanced by sources that are less numerous but more reliable (sources m) (figure 35).

IRIT-UPS