QoE-Driven Centralized Scheduling for HTTP
Adaptive Video Streaming Transmission over
Wireless Networks
Tiantian Li
∗
, Haixia Zhang
∗
, Jie Tian
†
, Shuaishuai Guo
∗
∗
Shandong Provincial Key Laboratory of Wireless Communication Technologies, Shandong University, Jinan, China, 250100
†
School of Information Science and Engineering, Shandong Normal University, Jinan, China, 250014
Email: tiantianli@mail.sdu.edu.cn, haixia.zhang@sdu.edu.cn
Abstract—In this paper, we propose a novel QoE-driven cen-
tralized scheduling framework to satisfy the differentiated quality
of experience (QoE) requirements of cellular users for HTTP-
based video streaming transmission. The framework jointly
considers the comprehensive information from the video server
side (video coding rates and QoE parameters), the user side
(video playback buffer occupancy), and the time-varying wireless
channel side (channel state information-CSI). A mixed binary
integer programming problem is formulated to maximize the
overall user QoE through the effective weighted resource allo-
cation and video coding rate adaptation, which is implemented
by the centralized scheduler residing at the Base Station (BS).
Our heuristic algorithm can be considered as a proactive one
that enables the proxy at the BS to rewrite the HTTP requests.
In particular, the impact of delay factor on the performance
of average rebuffering time is analyzed. Simulation results
demonstrate that our proposed scheme performs effectively in
enhancing the user QoE satisfaction and outperforms the four
benchmarks.
I. INTRODUCTION
With the proliferation of intelligent mobile phones and
tablets, people are more inclined to watch videos via mo-
bile devices benefiting from the great convenience to car-
ry them. According to historical data, the explosive video
traffic has accounted for 70% approximately of the whole
Internet traffic. Meanwhile, the ever-increasing mobile video-
on-demand (VoD) traffic poses tremendous challenges on the
congested wireless networks [1]. How to satisfy the stringent
requirements on conditions of bandwidth-limited and power-
constrained wireless networks becomes an essential problem.
The conventional approach to satisfy the video requirements
mainly concentrates on improving the quality of service (QoS)
of networks (i.e., broaden network bandwidth, enhance trans-
mission rate, degrade network congestion or delay) [2], [3].
Recently, the significant shift from ’connection-centered’ to
’user-centered’ networks simultaneously enables the transition
of evaluation criterion from QoS to QoE, which comprehen-
sively reflects the actual feelings or service satisfaction from
the user perspectives. It seems appealing for network operators
to enhance video QoE, i.e., high quality (definition), fluent
viewing (without interruption) and stable quality (without
quality fluctuation). There have been some literatures on QoE
optimization for video streaming [4], [5]. Nevertheless, the
QoE model based on mean opinion score (MOS) in these
works still keeps far away from the real-time user experience
during viewing. Consequently, a more comprehensive QoE
model is investigated in [6], but they fail to tackle the dif-
ferentiated QoE requirements when confronted with different
user equipments (UE), video types or user preferences.
Different from the Real-Time Transport Protocol/User Data-
gram Protocol (RTP/UDP) based streaming usually dedicated
to real-time video conference or live broadcast due to its lower
latency [7], HTTP-based streaming utilizes TCP/IP to ensure
high transmission reliability for VoD users and allows the
associated providers to reuse existing network infrastructures
and web servers that can ensure users to easily access services
across firewall without exploring new application protocols.
Due to these mentioned benefits, numerous providers are
scrambling to deploy HTTP-based Adaptive Streaming (HAS).
In HAS, a video can be divided into multiple independent
segments, each with fixed playback duration. The video server
encodes each video into several versions with different coding
rate (quality) before transmission. UE can employ an adaptive
coding rate selection algorithm to choose the most suitable
quality version for each segment to match the time-varying
wireless networks by sending HTTP request. E.g., at the user
side, [8] investigates the throughput-aware adaptive algorith-
m (TA) to match the channel throughput; [9] adopt video
buffer-aware adaptive algorithm (BA) to avoid occurrence of
interruption events, buffer overflow or unnecessary download
incurred by sudden stop; [10] integrates the two algorithms
mentioned above (TBA). However, these works specific to a
single client are one-sided on QoE enhancement, the adopted
algorithms only passively adjust the video rate to adapt the
network dynamics. To address these issues, [6] focuses on
improving QoE performance by resource allocation at the BS
side, then each user reacts to the allocation. But this scheme
is reactive for users when choosing segment rate according to
previous network throughput failing to reflect current CSI.
Motivated by the above discussions, we present a novel
QoE-driven centralized scheduling framework taking different
properties of UE and video types into consideration. By
taking full advantage of the comprehensive information, a
non-convex problem is formulated to maximize the overall
user QoE through weighted resource allocation and video
coding rate adaptation, respectively. Our heuristic algorithm