Page 76 - Kaleidoscope Academic Conference Proceedings 2021

2021 ITU Kaleidoscope Academic Conference

large to boost overall system capacity. T last is initialized priority-based adaptive preemptive scheduler are further
to a minor value i.e. 10 −5 in our simulation. highlighted as follows.
• The proposed scheduler pre-orders and enables HP
The priority-based adaptive scheduler is innovative and
UE to preempt the sub-bands according to their CQI in
unique in the following aspects:
order to deliver the service with requirements met.
• The sub-bands should be assigned, in the first place,
• The proposed scheduler adaptively tunes the priority
to a predefined set of high priority UE considering their
related weight in the sub-band allocation metric
individual channel status and buffer size.
calculation intending to minimize impact on the
• The remaining sub-bands should be assigned
throughput of LP UE.
considering the priority impact in addition to the
channel status differentiation between different UEs.
4. EVALUATION RESULTS FOR
Algorithm 2 The procedure of priority-based adaptive
PRIORITY-BASED ADAPTIVE
preemptive scheduler
PREEMPTIVE SCHEDULER
Input: u, p, s, N SB , N RB , v, T last , α
1: Find the scheduled high priority user set u HP := In this section, we provide system-level simulation
{i HP } and the scheduled low priority user set results using the simulation parameters in Table 2 to
u LP := {i LP } based on the service requested by assess the performance gain of the proposed scheduler.
the UE in u. A Single-User (SU) MIMO scheduler is assumed to
2: Use mapping function F rbir (sinr) to map the CQI eliminate user pairing and thus better visualize the
(j), j = 1, 2, 3, · · · , N SB for any high
s i HP (j) to r i HP preemption influence on overall capacity. The traffic of
priority UE i HP within u HP . The MIR across all LP UE is taken from one typical XR traffic from the
.
the sub-bands are denoted as r i HP ongoing 3GPP XR study item while that of HP UE
3: Calculate the sub-band level instantaneous mean assimilates typical uRLLC traffic. For simplicity, we
throughput T i HP for any high priority UE i HP consider the application data unit consists of a single
within u HP frame. This assumption is also in line with the current
NR 5G QoS identifier value, where packet level priority
1 N SB is still not available within a given stream.
X
= (j) ∗ N RB ∗ v. (11)
T i HP r i HP
N SB
j=1 Following the evaluation methodology of the 3GPP
XR study item, capacity is defined as the number
∗
of UEs among which 90 percent should be satisfied
4: Find the high priority user i with maximum T i HP
and order the sub-bands according to reported CQI for the transmission. Thus, we can see that the
∗
of user i . Then, let d SB denote the indexes of
capacity is 9 UE with the proposed scheduler and 8
sub-bands after ordering.
UE for the conventional proportional fairness scheduler.
5: for Sub-band j in d SB do The performance gain is thus 12.5 percent. Note
∗
6: if Data amount of user i is not 0 then that performance gain is expected to increase under
∗
7: Schedule user i in the j-th sub-band.
the circumstances of multi-streams XR traffic wherein
8: else the overlapping of high priority FoV tiles occupying
∗
9: for user i 6= i in u do
relatively much fewer time frequency resources and low
10: Use mapping function F rbir (sinr) to map priority non-FoV frames occur more frequently. Further
the s i to r i for all users waiting for fine-tuning on the HP UE set based on jointly the
scheduling. buffer status and the channel quality as well as the
11: Calculate instantaneous throughput priority related weight should as well boost the capacity
j
T according to Eq. (7).
i,ins performance. Moreover, if a finer preemption pattern
j
12: Calculate average throughput T beyond sub-band and the rate matching process of
i,avr
according to Eq. (10). XR can take the preemption pattern into account and
j
13: Calculate the metric M for each user in leave out in advance the traffic allocated for concurrent
i
the j-th sub-band according to Eq. (8) uRLLC services, the performance gain shall further
and Eq. (9). boost as well.
14: end for
15: Find the user with maximum metric and 5. CONCLUSION
schedule the user in the j-th sub-band.
16: end if This paper proposes a novel priority-based adaptive
17: end for preemptive scheduler which is promising to address
the capacity challenge of both multi-streams XR
Algorithm 2 synthesizes an illustrative description of services or XR services concurrent with other services.
the above two bullets with the predefined high priority Further capacity enhancement can be anticipated with
UE set reduced to a single UE. The advantages of some cross-layer specification adaptation to deliver
finer preemption pattern or adaptive rate matching
– 14 –