Comyco: Quality-Aware Adaptive Video Streaming via Imitation Learning

Abstract

Learning-based Adaptive Bit Rate (ABR) method, aiming to learn outstanding strategies without any presumptions, has become one of the research hotspots for adaptive streaming. However, it is still suffering from several issues, i.e., low sample efficiency and lack of awareness of the video quality information. In this paper, we propose Comyco, a video quality-aware ABR approach that enormously improves the learning-based methods by tackling the above issues. Comyco trains the policy via imitating expert trajectories given by the instant solver, which can not only avoid redundant exploration but also make better use of the collected samples. Meanwhile, Comyco attempts to pick the chunk with higher perceptual video qualities rather than video bitrates. To achieve this, we construct Comyco's neural network architecture, video datasets and QoE metrics with video quality features. Using trace-driven and real world experiments, we demonstrate significant improvements of Comyco's sample efficiency in comparison to prior work, with 1700x improvements in terms of the number of samples required and 16x improvements on training time required. Moreover, results illustrate that Comyco outperforms previously proposed methods, with the improvements on average QoE of 7.5% -16.79%. Especially, Comyco also surpasses state-of-the-art approach Pensieve by 7.37% on average video quality under the same rebuffering time.

A Time-Variant QoE Model Based on Real Video Streaming Data

“…We split this database in a ratio of 80/20 for fitting and testing. Four existing representative QoE models are selected as baselines for comparison: QoE lin and QoE log , with q(R n ) = R n and q(R n ) = log(R n /R MIN ), respectively; QoE vqa , where the VQA metric VMAF [12] is used for q(R n ); Comyco, the model used in [13], where up-and down-switches of bitrate are weighted separately.…”

Video Quality of Experience Metric for Dynamic Adaptive Streaming Services Using DASH Standard and Deep Spatial-Temporal Representation of Video

“…This model is found to be highly accurate. In recent years, many works [12,13] have used the QoE prediction model established by machine learning to guide the client's bitrate selection in dynamic adaptive streaming services.…”

Preference-Aware Dynamic Bitrate Adaptation for Mobile Short-Form Video Feed Streaming

“…For example, buffer-based (BB) rate adaptation [8] is a typical QoEinsensitive strategy to avoid rebuffering based on the mobile devices' playback buffer condition.Regarding the QoEsensitive video streaming strategies, theoretical control, such as MPC [5], convex optimization, such as BOLA [6], and RL techniques, such as Pensieve [7], are three most common methods which can work well in specific scenarios. Recently deep learning methods [9]- [11] have also been leveraged to improve the performance of adpative video streaming. Beyond that, much efforts have been made on implementing collaborative information to facilitate video streaming, e.g., CS2P [12] which focuses on improving the wireless network bandwidth estimation for bitrate adaptation, SDNDASH [13] which introduces central control to coordinate the usage of the limited bandwidth resources.…”

“…It first trains the parameters of preference prediction module based on the user behaviors (line 2). Afterward, the model repeatedly simulates the video playback events and optimizes the bitrate allocation model (line [3][4][5][6][7][8][9][10][11][12][13][14][15][16][17]. The parameters are trained based on the Q Actor-Critic algorithm and eventually, return the optimal policy π * θ .…”