1304 IEEE TRANSACTIONS ON MULTIMEDIA, VOL. 21, NO. 5, MAY 2019
multi-objective combinational optimization problem, subject to
the viewing-delay constraint. More specifically, the transcod-
ing chunk quality, the modulation and channel coding schemes
(MCS) for different links, and the viewport rendering offload-
ing determination parameter are calculated to maximize the
viewport quality, while minimizing the energy consumption of
the UE.
The rest of the paper is organized as follows. Section II pro-
vides a detailed analysis of the related work. In Section III, the
system model of the proposed MEC-assisted PVRV streaming
scheme over MC-based mmWave network is presented. The
problem formulation and our solution approach are described in
Section IV. Section V presents the simulation results. Finally,
Section VI concludes the paper.
II. R
ELATED WORK
A. PVRV Delivery
The spatial resolution of PVRV is considerably larger
than HD video. Consequently, the delivery of high-quality
PVRV to the mobile device is very challenging due to the
bandwidth-constrained network. To save bandwidth, one could
use high-efficiency compression techniques, e.g., H.264/AVC-
based PVRV coding [10], and H.265/HEVC-based PVRV
coding [11]. Since compression is closely related to data repre-
sentation, the studies in [12] and [13] have proposed content-
adaptive or viewport-centric-adaptive representations that of-
fer flexible delivery options. To facilitate interactive viewport
streaming, a tile-based/strip-based encoding scheme was pro-
posed in [14]. For interactive PVRV streaming, only tiles within
the field of view ( FOV) are delivered to the end-user. Consid-
ering dynamic FOV, a viewport-dependent adaptive streaming
system was proposed in [15] to reduce the bandwidth require-
ments. By exploiting the advantages of HTTP-based adaptive
streaming, tiling-based PVRV streaming over DASH was also
proposed to optimize QoE [16]–[18].
The previous studies are aimed primarily at wireline PVRV
transmission or offline delivery with local viewing on the device.
Current PVRV delivery mechanisms lack an optimized deliv-
ery that is tailored to wireless networks. Since mobile network
capacity is gradually growing, it can accommodate multi-user
PVRV traffic and lead to a new era of video communication.
However, traditional mobile streaming schemes are not very
efficient for PVRV due to the high bitrate of PVRV. Conse-
quently, more sophisticated PVRV streaming schemes have to
be designed for emerging mobile wireless networks.
B. Video Communication Over mmWave
Recent advances in mmWave wireless communication allow
for the support of high throughput applications. Preliminary
results indicate that a high data-rate mmWave link can commu-
nicate ultra-low delay and uncompressed video traffic without
problems [19]. However, one of the key challenges of mmWave
communication is channel dynamics: The channel varies rapidly
and the link maybe completely broken due to non-LOS paths
between the transmitter and the receiver. For video traffic over
mmWave, the authors in [20] proposed to use a receiver buffer
and data transmission scheduling for video bitrate adaption that
matches the channel. In another work [8], an MC-based LTE-5G
integrated architecture was proposed to improve the reliability
of mmWave communication but not for video traffic. Classic
techniques for low-delay video streaming with unequal error
protection were proposed in [21] but in this case for a 60 GHz
mmWave link. Similarly, by adopting network coding to en-
hance the packet transmission reliability, the work in [45] con-
firmed the feasibility of streaming video over cellular MC-based
mmWave links.
Furthermore, two recent works have proposed the delivery
of virtual reality game videos over mmWave networks. To cope
with the problem of PVRV signal intermittence caused by block-
ages, the authors in [22] proposed to add a mmWave mirror
device to relay the blocked signal, but they ignored the signal
blockage problem of the added mirror device. In [23], the en-
ergy efficiency of a mmWave BS was optimized with dynamic
power allocation for PVRV transmission.
The previously mentioned techniques exploit to the fullest
mmWave communication for ordinary video traffic. However,
they are not directly applicable to PVRV streaming since PVRV
traffic has certain specific characteristics. Though the recent
works have tried to optimize PVRV applications over mmWave
networks, they were still focused on the mmWave network opti-
mization and were not concerned with the particular characteris-
tics of PVRV traffic. It is still unclear how PVRV streaming per-
formance is related to the dynamics of an MC-based mmWave
channel, when it is affected by image tiling and viewport
changing.
C. Mobile Edge Computing
Modern mobile applications are sophisticated and usually
have a high computational load. This is unacceptable for a
power-limited mobile device. The rise of MEC makes up for
the inability of mobile devices to dedicate computing resources
for a protracted time period [24]. Mobile computation on the
UE can be partially transferred/offloaded to the MEC server.
Even though there have been studies on computational offload-
ing in mobile networks, they did not focus on the internals of the
communication system during computational offloading. Within
the communication subsystem, link adaptation and video chunk
quality selection can be jointly optimized with a trade-off be-
tween communication and computational offloading to further
improve the system performance. Additionally, when the MEC
architecture is extended to PVRV streaming, the system opti-
mization problem needs to consider adaptive viewport rendering
offloading.
III. S
YSTEM MODEL
The proposed PVRV streaming system consists of three key
elements: the content provider, the MEC server and the MC-
based mmWave/sub-6 GHz cellular network. We provide a brief
overview of the overall system functionality in the first subsec-
tion. In the remaining subsections we discuss the necessary
mathematical models for transcoding, the wireless links, and
Authorized licensed use limited to: Ningbo University. Downloaded on March 02,2021 at 06:52:41 UTC from IEEE Xplore. Restrictions apply.