Building a model of network interaction between the components of a multiagent system of mobile robots

• Main Authors: Vitalii Diduk, Valerii Hrytsenko, Andrii Yeromenko
• Format: Article
• Language: English
• Published: PC Technology Center 2020-10-01
• Series: Eastern-European Journal of Enterprise Technologies
• Subjects: multi-agent systems; mobile robots; machine learning; network model; web interface; websocket
• Online Access: http://journals.uran.ua/eejet/article/view/213989

The results reported here represent the first stage in the development of a full-featured laboratory system aimed at studying machine learning algorithms. The relevance of the current work is predetermined by the lack of network small-size mobile robots and appropriate control software that would make it possible to conduct field experiments in real time. This paper reports the selection of network data transmission technology for managing mobile robots in real time. Based on the chosen data transmission protocol, a complete stack of technologies of the network model of a multi-agent system of mobile robots has been proposed. This has made it possible to build a network model of the system that visualizes and investigates machine learning algorithms. In accordance with the requirements set by the OSI network model for constructing such systems, the model includes the following levels: 1) the lower level of data collection and controlling elements – mobile robots; 2) the top level of the model includes a user interface server and a business logic support server. Based on the built diagram of the protocol stack and the network model, the software and hardware implementation of the obtained results has been carried out. This paper employed the JavaScript library React with a SPA technology (Single Page Application), a Virtual DOM technology (Document Object Model), stored in the device's RAM and synchronized with the actual DOM. That has made it possible to simplify the process of control over the clients and reduce network traffic. The model provides the opportunity to: 1) manage the prototypes of robot clients in real time; 2) reduce the use of network traffic, compared to other data transmission technologies; 3) reduce the load on the CPU processors of robots and servers;  4) virtually simulate an experiment; 5) investigate the implementation of machine learning algorithms