Rate Adaptation-based Real-time Video Transmission Over UAVs Networks

In recent years, UAVs(Unmanned Aerial Vehicles) are widely used in both military and research fields. UAVs networks as an application forms is the most proper one for specific missions, such as battlefield surveillance and disaster relief. Among all the data collected, real-time video streaming is especially valuable, because it provides users with explicit and intuitive descriptions of the critical area. In this paper, we analyze the possible challenge lies in implementing real-time streaming over UAVs networks. Furthermore, we points out the problems of prediction inaccurate, ineffective strategy and lack of support for special network occasions within existing works utilizing bit-rate adaptation.Therefore, our research work aims at designing an effective bit-rate adaptation algorithm,which would be carried out in the video transmission process over UAVs network. The algorithm involves two main key criterion, maintaining continuous video playback and providing higher quality.First, we study the pattern of UAV trajectory in a specific mission scenario. During the mission, the waypoints of an UAV are predefined and the UAV repeatedly cruises in the same reconnaissance route, which makes a proactive train stage possible and meaningful.We propose to perform TCP probing in the train stage, and utilize time series analysis method to analyze the measurements. Results show that the available throughput varies periodically with UAVs' mission-related movements, and a raw throughput prediction version is thus calculated based on the knowledge of periodicity.Second, we use a queue model to analyze the fluctuation of video buffer and points that time as a parameter is more proper to reflect the situation of the buffer. Then we propose a novel method for the estimation of buffered video time, considering the diversity of received video segments' bit-rate during a transmission interval. Introducing the estimation value as a reactive control parameter, the raw prediction from train stage is further calibrated dynamically to adapt to the real-time environment.Third, we design an adaptation algorithm based on prediction to dynamically match the time-varying available throughput. The algorithm aggressively switch down bit-rate to maintain real-time performance, and selectively switch up for a better quality with smooth adjustment behavior. Furthermore, we novelly propose a special mode named slider mode to cope with the common low-bandwidth occasions in multi-hop UAVs network.The proposed scheme and logic are validated in simulation. Simulation results show that our throughput prediction scheme exhibits higher accuracy, and our adaptation scheme successfully maintains a continuous playback with a high quality and shorten the start-up delay compared with two constant bit-rate schemes. Also, we implement the slider mode in real environment using VLC, and introduce real-time video transmission application over UAVs networks into the ground control station platform we build.