Research of Video Stream Frame Delay in UAV FPV-Control Information Exchange Channel in Hybrid Communication Network Terrestrial Segment

This research considers the dependence of the delay and loss of video stream frames compressed by neural network codec developed on the basis of neural network variation auto-encoder on the size of transmitted frames in the realization of information exchange channels between unmanned aviation system and external pilot station in the ground segment of hybrid orbital-terrestrial communication network taking into account the distance between them when using 3G and LTE data transmission technologies are used. The Relevance of the research is conditioned by the necessity to achieve a given level of service quality for FPV control of UAVs in communication networks. Methods used. In this research, the applied transmission delays and frame loss of FPV control video stream when using neural network codecs are measured by in-situ experiment. The applied delays and losses take into account segmentation, packet recovery and transmission of multiple UDP packets for each payload. Additionally, the Rosenblatt-Parzen method reconstructs the probability density distributions of delay probabilities. Results. Estimates of average values of transmission delay and frame loss of video stream (compressed by neural network codec) when using 3G and LTE data transmission technologies taking into account different distances between the unmanned system and the external pilot's station are obtained. The distributions of video stream delay dependencies on the payload size are reconstructed. The character of video stream delay distribution formed by neural network codec is found. The Novelty of the obtained results lies in the study of the nature of delays and frame losses of the FPV-control service video stream transmitted through mobile communication networks at the application layer of the OSI model when using neural network codecs. Practical significance. The results can be used in modeling of information exchange channels for FPV control in order to form the optimal configuration of the used neural network codecs.