5G Toolkit: Steering 5G New Infrastructure into Fast Lane

 

With the announcement of the results of the second phase of 5G bidding by China's three major operators at the beginning of 2020, large-scale 5G deployment has been carried out across the country. The scale of nearly 500,000 base stations has quickly made China a hot spot for 5G networks. It has become the focus of operators to implement rapid 5G site commissioning, optimize basic coverage and establish market competitiveness. To meet the wave of 5G new infrastructure, ZTE has made plans in advance and launched a variety of intelligent tools for rapid deployment and optimization of quality 5G networks.

SON: Self-Configuration Self-Optimization for High Efficiency
Configuration and optimization of neighbor cells, ENDC X2/Xn, and PCI are the first step in 5G site commissioning. However, inefficient manual operation is very difficult to meet the requirements of large-scale site commissioning. ZTE has developed 5G self-organizing network (SON) that uses AI algorithm to identify problems, improve maintenance efficiency, and enrich maintenance means. It realizes intelligent identification, organization, orchestration and error correction of neighbor cells and links, and thus facilitates rapid 5G deployment. 
When a 5G site is powered on and its links are set up, the SON module of UME monitors the link setup and self-configures its neighbor cells according to the imported neighbor cell planning table. After the self-configuration of neighbor cells is completed, the system will self-check whether ENDC X2/Xn interface links are established between 4G and 5G sites. If not, the system will automatically initiate the link establishment without manual operation. SON can self-optimize neighbor cell configuration errors as well as PCI conflicts and confusions that may exist in the network. At present, since a 5G terminal chip does not support 5G NCGI measurement, ZTE adopts the solution of discovering unknown PCI and engineering parameters via the air interface to determine NCGI of the target neighbor cell. Reasonable and accurate configuration of neighbor cells can be ensured by setting the distance threshold, the RSRP threshold of unknown PCI measurement, and the number of unknown PCI measurements. Based on the configuration of X2/Xn interfaces and neighbor cells, the system detects PCI conflicts and confusions. If there is any PCI conflict, the system will self-optimize the PCI conflict and reallocate PCIs for 5G-5G and 4G-5G neighbor cells. 
At the 5G deployment site in Fujian, SON completes the commissioning of 100 sites in three miuites compared with 20 minutes for a single site in the traditional mode. Automatic monitoring ensures that data is configured correctly and the efficiency is increased a hundred times.

WNG: Automatic Drive Test for Easy Site Acceptance 
Large-scale site commissioning will inevitably bring a lot of drive test acceptance. Traditionally, a single site acceptance test needs three persons in a group who carry a laptop, a GPS, a frequency sweeper, MOS meter and test terminal to collect data during the day and analyze data overtime at night. The testers have a large workload and low efficiency, which affects site acceptance and access to the network. 
Wireless network guardian (WNG), a new tool in the 5G toolkit supplied by ZTE, is an automatic drive test solution based on cloud server and APP architecture (Fig. 1). It can automate the whole process of wireless data collection and network evaluation and analysis report output, making the site acceptance easy.

 

The WNG service analysis and processing system is deployed on the cloud server. Professionals can set test tasks remotely by logging in to the server through the client. The test items include parameter verification and cell performance verification. The system can provide standard templates, customize test items, and perform multi-task distribution. Now a site acceptance test can be easily started only with one person, one vehicle and one terminal. After receiving the task, the terminal automatically collects data and reports it in real time. The server automatically analyzes and processes the data, and outputs a test report after the test is completed. The report template can also be customized as required. Testers don't have to stay up late to make reports anymore.
In China, more than 50 outfields have applied WNG as a single test tool for 5G sites. More 500 terminals are used in test task and more than 1000 reports are output every week. Compared with the traditional drive test mode, the number of personnel has been reduced by 60%, and the efficiency has been improved by 65%. 

AAPC: Automatic Antenna Pattern Control for Effective Coverage Enhancement
Massive MIMO, one of the key 5G technologies, can effectively improve three-dimensional extended coverage and system capacity in complex scenarios by using large-scale array antennas and three-dimensional beamforming. Compared with the traditional antenna, Massive MIMO has more parameter adjustment dimensions, including horizontal beam-width, vertical beam-width, azimuth, down-tilt angle and the number of beams. Each dimension can be well adjusted by setting a reasonable step size. Theoretically, there are tens of thousands of possible antenna parameter weights for a cell. In a actual 5G network, it is almost impossible to perform manual multi-cell collaborative optimization and adjustment according to changes in the scenarios and services.
ZTE's automatic antenna pattern control (AAPC) adopts AI-based ant colony algorithm to optimize antenna weight groups based on the balance between optimization target and search time. Through simulation learning, the weight group with no obvious effect is abandoned. AAPC sets the maximum number of iterations to reduce the requirement of computing power and greatly reduce the search time for an optimal solution. It gives optimal weights according to different scenarios, user distribution and optimization targets. After the optimal weights are executed by NE, the optimization data is reported by UE to verify the effect, and the next optimization iteration will be performed. In case of KPI deterioration, the previous weight group will be returned directly. In field trials, the test verified that it used to take more than 40 man-days to manually optimize antenna weights, but now it takes only two man-days to complete the optimization using AAPC, with high efficiency and good effect. 

AI: Smart KPI Detection for Early Troubleshooting
KPI is a direct reflection of network quality. An KPI anomaly means that some failure may occur in the network. O&M engineers have to deal with hundreds of KPI changes and alarms every day. Many times they are exhausted and it is difficult for them to monitor KPIs accurately and quickly. It is often not until users complain that they find KPI faults and start the handling process. With the help of AI new technologies, ZTE's wireless intelligent O&M system combines machine learning (ML) and expert rules to implement automatic KPI anomaly detection and fault diagnosis in the network, which is equivalent to a 24-hour “network health monitoring and diagnosis instrument” that provides network doctors with analytical data and root cause diagnosis. The system introduces a time series clustering method based on structure feature to classify KPIs. For each KPI type, the appropriate time series model is selected to predict its normal baseline in the next time granularity. If the real-time measurement value of KPIs exceeds the baseline, it can be found in the first time granularity. 
When the system detects a KPI fault such as a defined known fault, the system will make a comprehensive analysis based on the correlated alarms, operation logs, network topology and expert rule base, and give root cause judgement and troubleshooting suggestions (Fig. 2). When an unknown fault is detected, the ML-based diagnosis module will use the partial least square (PLS) regression algorithm to analyze and locate the root cause. Through the contribution analysis of possible causes, the root counter index at the top is found as the abnormal root cause. 

 

During the field verification, when the success rate of LTE E-RAB setup of a subnet suddenly drops from 99.9% to 99.2%, the system will soon discover this anomaly and drill down to analyze the root cause and rapidly locate the eNodeB of ID 208203. It is found that the success rate index of a cell in the eNodeB drops to 0, which makes the corresponding network indexes fluctuate abnormally. Through the alarm correlation analysis, when the KPI is abnormal, the RRU in the cell fails and is out of service. It is concluded that the RRU fault leads to the abnormal KPI change, and the time for the whole fault location analysis is less than 10 minutes. But through manual analysis, it took an experienced engineer at least two hours to analyze KPIs through the network management system to correlate alarms and logs, and drill down to the TopN cell to find out the problem. 

In the new infrastructure construction, 5G will become a new artery of social information flow and an accelerator for industrial transformation and economic development. It is also the common goal of operators and partners to build a 5G network with perfect coverage, high quality and leading services. Focusing on new infrastructure construction and long-term quality network operation, ZTE's 5G toolkit introduces new AI technologies to analyze the pain points encountered in network deployment, operation and optimization, improve the efficiency of 5G site commissioning and quickly optimize the network. The 5G toolkit is developed to steer 5G new infrastructure into the fast lane and help countries achieve their economic and strategic transformation.