Smooth DASH adaptation exploiting throughput prediction

We present and evaluate a procedure that exploits throughput prediction to select the DASH video quality sequence (sequence of representations) with the highest average bit rate and the fewest quality switches, while explicitly taking into account throughput and time uncertainty. Experiments with an Android implementation of the proposed procedure show that it can exploit throughput prediction to achieve a high QoE (Quality of Experience) with a small number of video pauses and video quality switches, even when the throughput and time uncertainty is large.