Page 74 - Kaleidoscope Academic Conference Proceedings 2021

2021 ITU Kaleidoscope Academic Conference

Table 1 – Summary of mean packet size ratio for I-frame and P-frame

Configuration I-frame mean packet size (Byte) P-frame mean packet size (Byte) Ratio
VR2-1 5991.3 3218.7 1.86
VR2-3 6328.7 3318.4 1.91
VR2-8 8374.2 4623.4 1.81

unit could be undermined in case the packet should
account for the highest importance amongst all the
packets therein due to either the inherent importance
of the packet itself or the decoding situation of the
application data unit. For example, if the packet
itself is the final component of the application data
unit whose preceding packets have been successfully
decoded, over-preempting due to uRLLC services and
the consequent decoding failure of the packet and
invalidation of all the previous successfully transmitted
packets. To address this, the following alternatives
could be considered to better support XR performance
Figure 2 – The structure of tile-based FoV/non-FoV in case of concurrent transmission.
streaming • Enhanced scheduling considering the service QoS
including the priority information of a packet.
3. ON BETTER SUPPORT OF XR AND • Specification adaptation to facilitate the scheduling
CONCURRENT SERVICES e.g. on a finer granularity or using an exact rate
matching pattern.
Given the aforementioned XR traffic characteristics, the
The above considerations are suitable to semi-statically
challenge is two fold:
scheduled uRLLC services and dynamically scheduled
• The data rate requirement which can be translated
XR services overlapping as well.
into a huge transport block in a jittering arrival manner.
• The stringent QoS requirement of reliability and
latency.
For the first bullet, grant-based scheduling is more
suitable given the jitter impact can be properly handled
under this mechanism.
For the second bullet, sufficient radio resources including
time, frequency and spatial domain resources should be
leveraged to ensure the successful delivery of the file
within the latency and reliability budget.
Under the scenario that a grant-based uRLLC is
requiring overlapping time frequency resources as the
grant-based XR service, uRLLC with potentially a
higher priority than a scheduled XR packet is likely Figure 3 – The illustration of preemption.
to preempt the XR service as illustrated in Figure 3.
Nevertheless, according to the current Downlink (DL)
preemption mechanism, the time frequency resources 3.1 Priority-based adaptive preemptive
determined by the periodicity of search space set of scheduler
DL Control Indicator (DCI) format 2_1 and the DL
active Bandwidth Part (BWP) are equally divided into DL preemption was first introduced in 3GPP Release-15
14 grids either consisting of all the activated BWPs to handle the coexistence of uRLLC service and
plus a time domain basic unit or half of the frequency concurrent eMBB services. Normally for grant-based
domain activated BWPs plus two time domain basic uRLLC in particular for retransmission, it’s pretty hard
units. For XR services featuring tight delay budget and to avoid already scheduled eMBB transmission due
resultant limited if not unavailable at all retransmission to the air interface delay budget. In that case, an
opportunities, similar preemption shown in Figure 3 additional indication can be sent to UE to facilitate
could be disastrous and lead to the decoding failure of decoding by neglecting the preempted part over the
the XR packet. Worse still, the whole application data resource reserved initially for eMBB. The preemption

– 12 –