Test Cases

To be able to compare the different methods we will use a set of test cases as common input to all approaches. The criteria for these test cases are derived from basic properties of the above introduced experimental scenario.

WORLD	MEASUREMENT	AGENT
Number of objects $n_{obj}$	Frequency $f_{meas}$	Frequency $f_{perc}$
Number pos/obj $n_{pos_obj}$	Number pos/meas $n_{pos_meas}$	Number pos/perc $n_{pos_perc}$
static/ dynamic objects $s or d$	--	Processing capacity $cap_{ag_proc}$

It makes a difference wether the objects in the world are static objects and/or dynamic ones, or whether they occupy only one position or more than one. Furthermore it makes a big difference whether the frequency of measurement is to 'slow' for to get all important informations. Besides this it can be important whether ohnly one data point ist transmitted per measurement (called sequentially) or a whole array (called parallel). Finally It can make a difference how the frequency and capacity of the agent perception and/ or processing has to be classified. Using these criteria we have for each testcase a kind of test signature like $n_{obj}=3$, $s$, $n_{pos_obj}=4$, $n_{pos_meas}=1$, Sequ., saying that there are three static objects occupying four positions each; during one measurement there l only on position of the area be recorded and transmitted. Therefore this is a sequentially data stream. Generally it is assumed that the measured input data are assumed as n-dimensional vectors $x_{i} \in \Re^{n}$.