| Summary: | Recent Multirobot systems (MRS) are moving from theoretical considerations and from development and research centres to the area of practical applications. The solutions to real practical problems bring new challenges which are derived from actual requirements, while they are also interesting from theoretical point of view. One of the interesting areas of investigation is the problem of distributed data processing by limited computing and communication performance of individual components in MRS. In this article, the authors try to demonstrate, using a simple example, the possibilities of distributed solution of classification tasks. Such questions as: – To what extent it is appropriate to distribute the tasks among individual elements of the system and to what extent to minimalize the requests on the communication subsystem? – Is it more appropriate, in the design concept of distributed data processing, to use a data fusion system, features fusion or decision fusion? are not universally solvable. Therefore, we refrain from the analytical analysis and the choice of appropriate level of information fusion, but in four different scenarios we focus on solving ‘voice command recognition’ – we would like to show the advantages and disadvantages of individual approaches. The experiments described and the results achieved are based on simulation experiments and verified by experimental and demonstrative MRS.
|
|---|
