Data-Learning-based Uplink Intelligent AMC

Data-Learning-based Uplink Intelligent

AMC Feature/Parameter Description
Objective:
Uplink SINR prediction model building process
 To use Data-Learning-based Uplink Intelligent AMC to build the Uplink SINR prediction model .

Description:
 During uplink scheduling, after the eNodeB delivers an uplink grant to a UE, the UE sends uplink data several
TTIs later. This means that the MCS index indicated in the uplink grant delivered by the eNodeB may not
accurately match the uplink channel quality during uplink data transmission, affecting the uplink UE throughput .
 Data-learning-based uplink intelligent AMC collects a large amount of UE scheduling information, and uses the
collected data to train a neural network model. The collected information includes the uplink SINR at the time of
data transmission, uplink SINR measured in several historical channel measurements, scheduling interval, and
allocated RB positions. The trained model allows the eNodeB to predict the SINR at the uplink transmission
moment when delivering the uplink grant, thereby guiding the selection of the optimal uplink MCS index,
improving the transmission spectral efficiency of UEs, and increasing the uplink UE throughput.
 After a model is built, periodic monitoring of the model is started. If the difference between the predicted uplink
SINR and the actual uplink SINR exceeds the CellIntelAmcConfig.UlPredErrorUpperLimit parameter value,
data is collected again for model update.

Benefit:
 Uplink intelligent AMC increases the User Uplink Average Throughput.

Drawback:
 After uplink intelligent AMC is enabled, the CPU usage may increase.
 The average uplink MCS index and the uplink IBLER may change depending on UE distribution and the
proportion of bits of small packets.
Data-Learning-based Uplink Intelligent
AMC Proposed Change & Expected Results

Trial Changes:

Parameter MO Implementation Current Value Proposed

Value
UL_INTEL_AMC_SELECTION_SW: If this option is selected,
MCSs for uplink transmissions are determined based on
big data analysis. When the uplink intelligent AMC model
UL_INTEL_A
IntelligentAmcSwitch CELLINTELAMCCONFIG converges, the average uplink spectral efficiency of UEs
0 MC_SELECT
and average user-perceived uplink rate increase in the cell.
ION_SW-1
If this option is deselected, network performance is not
affected

Hardware:
 Base Station Models:
 3900 and 5900 series base stations.
 DBS3900 LampSite and DBS5900 LampSite.
 Boards:
 Main control board: The LMPT, UMPTa and UMPTb boards do not support this function.
 BBP: The LBBPc, LBBPd, and UBBPd do not support this function.

Expected Results:
 Increases the User Uplink Average Throughput.
 Data-Learning-based Uplink Intelligent
AMC Proposed Area/Cluster for the Trial
Cluster Selection:

It is good practice to enable uplink intelligent AMC when all of the following conditions are met:
 UEs are distributed near the cell center, at the cell edge, and at a medium distance from the cell center. The proportion of UEs whose
path loss falls within the range of [105, 135] exceeds 50%. ( The number of UEs whose path loss falls within the range of [105, 135] are
measured by L.Traffic.User.PL4 to L.Traffic.User.PL9 ) .
 The proportion of small-packet bits is 30–80%. ( Proportion of small-packet bits = L.Thrp.bits.UE.UL.SmallPkt/L.Thrp.bits.UL ) .
 The uplink PRB usage is greater than 10%. ( Uplink PRB usage = L.ChMeas.PRB.UL.Used.Avg/L.ChMeas.PRB.UL.Avail ) .

 Trail will be done on 15 sites on Central_Unaizah_City_Cluster_4

Intelligent Case Model Information 21, June 2022
Physical Site Name City Cluster
(EUZ007) Unaizah Central_Unaizah_City_Cluster_4
(EUZ012)_VIP Unaizah Central_Unaizah_City_Cluster_4
(EUZ019) Unaizah Central_Unaizah_City_Cluster_4
(EUZ020) Unaizah Central_Unaizah_City_Cluster_4
(EUZ031) Unaizah Central_Unaizah_City_Cluster_4
(EUZ032) Unaizah Central_Unaizah_City_Cluster_4
(EUZ037) Unaizah Central_Unaizah_City_Cluster_4
(EUZ047) Unaizah Central_Unaizah_City_Cluster_4
(EUZ964) Unaizah Central_Unaizah_City_Cluster_4
(EUZ967) Unaizah Central_Unaizah_City_Cluster_4
(EUZ972) Unaizah Central_Unaizah_City_Cluster_4
(EUZ974) Unaizah Central_Unaizah_City_Cluster_4
(EUZ975) Unaizah Central_Unaizah_City_Cluster_4
(EUZ981) Unaizah Central_Unaizah_City_Cluster_4
(EUZ996) Unaizah Central_Unaizah_City_Cluster_4
Data-Learning-based Uplink Intelligent
AMC 4G KPIs for the Trial
Data-Learning-based Uplink Intelligent
AMC 4G KPIs for the Trial
Data-Learning-based Uplink Intelligent
AMC 4G KPIs for the Trial
Data-Learning-based Uplink Intelligent AMC KPI Results

KPIs Pre (13-20 ) Post (21-28) Delta (+/-)

Observations: Total Traffic Volume_GB 19470.22 19684.69 1.10%
 UL User Average Throughput improved by 7.09 % 4G UL Traffic Volume_MB 1564997.65 1539413.93 -1.63%
 Average CQI improved by 0. 16%
 Huawei LTE Avg Users 4528.73 4485.59 -0.95%
Average UL IBLER improved by 5. 60%
 Average UL RBLER improved by 4. 02% 4G_H_UL User Avg throughput Mbps eRAN7 3.24 3.47 7.09%
 Average UL MCS improved by 0.7% 4G_H_Cell UL Avg Throughput Mbps_Eran7 3.53 3.46 -1.91%
 1.10% increase in Total 4G Data Traffic 4G_H_UL-PRB-Utilization 14.73% 14.46% -1.87%
 All other major KPIs are following the normal trend L.UL.Interference.Avg -114.18 -114.23 0.04%
4G_UL IBLER 8.73% 8.24% -5.60%
4G_H_Max Retrans BLER UL 0.142% 0.136% -4.02%
4G_H_UL Retransmission Rate 11.21% 10.82% -3.52%
4G Avg UL MCS 19.20 19.33 0.70%
4G_H_Avg_CQI 9.47 9.48 0.16%

 It is clearly from KPIs that there is a significant improvement in UL User throughput

Conclusion:  As there is no negative impact observed, it is suggested to rollout
Thank You!
‫ُش كرًا‬