SkyEyes: adaptive video streaming from UAVs

Abstract

UAVs (unmanned aerial vehicles) equipped with high-end cameras have become increasingly popular among consumers. UAVs have been traditionally considered for applications such as disaster response and surveillance, while emerging applications include live-event broadcasts, precision agriculture, and augmented-reality games. Consumer UAVs today use fixed-bitrate video streaming where users configure the resolution (4K or 1080p). However, applications with real-time streaming that deploy UAVs in the wild will require adaptive video streaming to tackle uncertain wireless link capacities and meet their video quality requirements. This paper is a first step toward the design of adaptive video streaming algorithms that can provide significant gains for UAV streaming. Our system SkyEyes leverages two novel aspects to aid UAV streaming: content-based compression and video rate adaptation based on location sensors and client buffer status. Our system prototype, while far from complete, exhibits promise -- we believe that adaptive video streaming is indeed crucial for real-time UAV applications