Architecture of Proactive Localization Service for Cyber-Physical System's Users

In this article, a microservice architecture of cyber-physical space was considered and in particular localization service was implemented. We propose an architecture of proactive localization service, which allows predicting the activity of the tracked object in cyber-physical space. To solve the position prediction problem, we tested machine learning methods and contrasted the results of the trained models. Three machine learning algorithms were tested (artificial neural network, random forest, and decision tree) with two different datasets. The reason of high/low prediction accuracy were identified. As a result of testing, the best result has neural network. In this case mean absolute error is 8.2 and 11.7 m respectively for dataset №1 and №2, while random forest has 13 and 14 m error. The architecture of the service was developed using containerization technologies and special tools for their deployment and management.