MOSCOW, March 29. /TASS/. Researchers from Belgorod State University (BelSU) developed an algorithm for neuro-fuzzy systems for controlling data transmission in wireless self-assembling networks, the BelSU press office said. The new insights could substantially lower data losses in networks for special operations used for instance in controlling sensitive sites.
"The experiments demonstrated that the application of the suggested algorithm ensures a notable decrease in the mean value of transmission time of data flows, as well as the minimization of retransmitting," said Konstantin Polshchikov, one of the project’s initiators, and Director of the Institute of Engineering Technologies and Natural Sciences at BelSU.
The basic concept of self-assembling networks has appeared only recently, and initially it has been combined with standard mechanisms of controlling data transmission traffic applied in conventional networks with predefined topology. The fixed topology implies the presence of one or several well-defined aspects: physical carrier lines (e.g. communication cables of Ethernet network); functional-logic structure; and territorial structure. This caused delays in the delivery of data packages and an undesirable loss of information.
"In a wireless self-assembling network, in contrast to conventional ones, each network node can act not only as receiver or transmitter of information but also as router or forwarder. As the network data have normally low acceptance rate, the reboots could take place where the network discards some data packages. Hence, we pursued an objective to improve the already existing algorithm of controlling the intensity of data transmission," commented Sergey Lazarev, Executive Project Manager and Vice-Director of Scientific and Innovative Activities at Institute of Engineering Technologies and Natural Sciences at BelSU.
Upon creating the new algorithm, a neuron network approach was applied which resulted in "training" the network for adaptive changes in the intensity of data transmission and balance of the load. The algorithm provides the measurements of current values of response time required for accepting the transmission of a data package, neuron-fuzzy forecasting of this value in the next cycle and then the calculation of temporal delay upon data retransmission based on the forecast.
"Initially, every neuron network is trained based on a data selection: it is provided with basic models of behavior and analysis, and after the training stage, it organizes the work in the frame of learned logic, " Lazarev recapped. "In this study, we trained the algorithm with various critical situations and then it learned how to meet the solutions by itself to change various parameters of a wireless network."
The model proposed by the scientists was tuned and studied by means of simulated experiments where the functioning of the wireless self-assembling network was tuned in order to provide a connection while working in emergency situations in hazardous building facilities. The results of the experiment proved the new algorithm to be very efficient in the decrease of the mean transmission time of data flows.
This was done in the framework of the Federal Target Program "R&D in Priority Areas for the Development of Science and Technology in Russia in 2014-2020". The results of the study were published in the recent issue of the International Journal of Applied Mathematics and Statistics. The project is targeted at the wireless self-assembling network for building and operation of sensitive sites where the stable connection to transmit voice and digital data is of particular importance.