Video streaming over HTTP with dynamic resource prediction

Stochastic Hybrid Model Predictive Control using Gaussian Processes for Systems with Piecewise Residual Dynamics

“…While the true residual dynamics in an environment might varying as a function of the workspace variables, training the models in ĝset to depend on them implies that they will over-fit to the specific environment where the samples are collected, thus harming generalization to unknown environments. As a result, Remark 4.2.1 holds practical merit and is not particularly limiting unless the residual models in g set are desired to be finetuned online to the new environment in a streaming-based [40,31,109] fashion (a relevant problem that is not considered in this thesis). By Bayes' Rule,…”

A priority-based adaptive scheme for multi-view live streaming over HTTP

“…However, when the segment duration is long (e.g., 10s), the preceding method is not suitable to represent the instant fluctuations of the bandwidth. Another prediction algorithm is based on download throughput samples [10], it estimates the available transmission bandwidth of the next slice according to the average HTTP/TCP throughput of last segment. We have improved this method by dividing the segment into very short intervals, the accuracy of this algorithm is improved.…”

“…So it seems necessary to select another segment in the same time interval if the instance bit rate of the active one is estimated to be higher than the available bandwidth. Thang et al propose a bit rate prediction in [10] and their experiments show that it can be easily computed by obtaining some sample segments. However, most online video streaming are not real-time video techniques, they are often willing to tolerate a playback delay of several seconds or minutes in exchange for a smaller bandwidth requirement.…”

“…Since there are many existing methods to implement bandwidth prediction, we have improved on a method that utilizes the download throughput samples and adopt it in the approach for its high accuracy [10,11]. When multiple video streams are requested synchronously in the same transmission cycle, all the requests will get their appropriate segments at once.…”