AI based application in telecom network

 In this article , we will focus on the application , where Machine Learning can be used , so that telecom industry can get benefit of  this new revolutionary and promising domain. As telecom industry is facing a huge challenge of continuous tech-revolution and customer demand , since now customer are not just happy with the service provided they also value  Quality of experience which end user experience.

From end user point of view they not only need service , but with good user experience as well. for eg. for end user it doesn't matter which modulation techniques or Carrier Aggeration. techniques network is using , for them what matter is how much improvement they are getting while watching a video lets say in you tube , in terms of buffering time . Solution provided by AI can not only increase end user experience , but it also has ability to perform optimization of resource , which are very beneficial for telecom industry , especially Radio resources, thus can help telecom industry to fulfill there expectations.

Artificial Intelligence is transforming the way the telecommunication industry operates. The adoption of AI technology based solutions has grown, especially to drive efficiencies and to fulfil the consumers' demand for contextualized experiences. According to Transparency Market Research, "the global market for artificial intelligence is estimated to post an impressive 36.1% CAGR between 2020 to 2024, rising to a valuation of US$3,061.35 billion by the end of 2024."

The concept of bringing intelligence in telecom network has been greatly expected in research areas , which are increasingly dynamic, heterogeneous, large-scale, and complex. For a long time, researchers have been deploying solutions by adapting AI techniques, to automate management and optimization of telecom network. Implementation of AI techniques in telecom network, especially with current Heterogenous network with consist of small cell, macro cells for better coverage , can show dramatical results, I solving complex problem like optimization of network, fault tolerance , cell planning etc. since it will be a very complex task to be performed by any human, so by leveraging power of AI model to deal with this complex task with AI techniques like Neural network, Machine learning, Reinforcement Learning, can help Mobile network operator to solve these above mentioned complex problems.

Now lets look into some of the potential use cases of Machine Learning , which telecom industry can take leverage of.

1) Self  Configuring Network.

2) Self Healing Network

3) Self Optimization network

4) Smart Handover management to improve customer Quality of experience.

5) Prediction of Fault in telecom network.

1. Self  Configuring Network.

During the last few years, the technology of SONs has experienced an explosive growth in its study. SONs are to effectuate substantial reductions of operation cost by diminishing human involvement. The essential idea of SONs is to integrate network planning, configuration, and optimization into a single and mostly automated process requiring minimal manual intervention. Our present telecom network is very complex because of different connected device and it assumes to get more complex in coming future , it will become more complex with introduction of 5G and future technology.

 

As shown in Fig.1 , in our current telecom network we have different type of cell and with this we have very high dynamics for user and its services , which leads to manage a huge number of parameter for different feature and services, which needs a better planning. Since with human effort it tooks lot of time and chances of error are aso high for such kind of planning , AI based techniques eg. Neural network , Genetic algorithm can be leveraged to perform this task of planning in much better way above mentioned techniques can be used for smart cell planning , Physical channel indicator and Radio resource assignment in a much better way in current heterogeneous telecom networks.

2. Self Healing Network

One of the most critical factor in current heterogenous network is to make network robust , at present current SON gives some such kind of functionality like activating fall back in case of fault etc. with us of AI techniques like Reinforcement learning  a real time autonomous reconfiguration can be performed, by observing user pattern and network condition without terminating mobile usage ,by adapting adaptive policy-based dynamic reconfiguration framework, by creating and updating policies dynamically in response to changing reconfiguration requirements, with the help of reinforcement learning Framework.

 

3. Self Optimization network

With current heterogenous network , it is expected to improve performance , because of different cell type in customer network i.e ( small cell, macro cell, picocell, femtocell different inbuild solution ). With so many different type of cell complexity of the network increases and hence to improve performance better optimization mechanism is needed. With AI techniques there is a huge potential for real time optimization of these kind of complex network, which can inturn enhance user experience.

Few methods which can help in self optimization.

a) Automatic coverage optimization and load balancing ,can be achieved by adjusting the antenna settings and thus shaping the radio coverage, and by adjusting the handover parameters to logically change the cell size. And this proposed self-optimization of antenna tilt and power ,  can be achieved with the help of fuzzy neural network optimization based on reinforcement learning frame work.

Research work for the same has been published in this research paper "S. Fan, H. Tian, and C. Sengul, {Self-optimization of coverage and capacity based on a fuzzy neural network with cooperative reinforcement learning,} EURASIP J. Wireless Commun. Netw., vol. 2014, no. 1,pp. 57:1-57:14, Apr. 2014."

b) Mobility Optimization is to eliminate unnecessary handover , with smart cell selection method by using AI techniques like Q- Learning algorithm of Reinforcement learning frame work , handover algorithm can be optimized more , so that now handover decision will be based on history of user experinec  while handover has been done  and  based on that cell will be selected where probability of user experience is high, rather than conventional method of handover which is triggered by events eg A3.

c) Link quality estimation must be always performed with a high reliability to facilitate a secure transmission with robustness in HetNets. Rather than conventional static link-quality aware routing metrics that adopt simple moving average estimators, bio-inspired estimator based on the neural network paradigm can be utilized to improve the effectiveness of link-quality estimation.

d) Virtical Handover (VHO) in HetNets plays an important role in fulfilling seamless mobile service when users cross different cells with different link layer technologies for RANs. Current VHO algorithms mainly focus on when to trigger and what connection QoS to improve, but neglect the synthetical consideration of all currently available candidate networks, where AI-based techniques can help to get optimal decisions on parameters by overall evaluation of the complicated conditions. For instance, the  VHO problem for user QoS enhancement and system performance improvement in HetNets, and proposes an adaptive parameter adjustment algorithm based on neural network model. 

4) Smart Handover management to improve customer Quality of experience

In current telecom network , the handover algorithm which is been used is based on standard event i.e  A3 event, and are mainly focused on the optimization of event trigger parameters, e.g., Hysteresis, Time-to-Trigger and Cell individual Offset, this approach has the problem that it just consider strongest signal for target cell selection before handover , but it did not consider user Quality of experience.

For example, in scenarios where the handover to the strongest neighbour cell is successful but, a while after the handover, the transmission is deeply affected, e.g., by the presence of an obstacle, SOTA

handover algorithms are likely to lead to a severe degradation of QoE, due to the unpredicted cell outage , this problem can be solved by using AI based techniques like feed forward Neural networkor deep Reinforcement learning, where handover will be done based on previous QOE for user.

 

Fig.3  2 level of neural network architecture to perform smart handover.

With the above proposed AI techniques of like neural network which is depicted in Fig. 3, consists

in the following: the source eNB gathers the time series of UE measurement reports before the handover, which contains the Reference Signal Received Power (RSRP) and Reference Signal Received Quality (RSRQ) of the source and neighbour eNBs. The eNB also collects the information on the QoE of

the user as a result of past handover decisions. In suggested AI techniques this QoE is quantified by two metrics, 1) the probability of successfully downloading a file and 2) the file download time for completed downloads. Therefore, in suggested Neural network it is suggested  to use a two level neural network model to estimate these metrics, as shown in Fig 3. At level 1, first neural network (NN1) is trained using UE measurements as input, and the past QoE in terms of download complete/not complete as output. On the other hand, at level 2, second neural network (NN2) is trained using only

those UE measurements as input, for which the file download was completed, and the file download time as output. Once the training is completed the AI based  handover algorithm of the source eNB uses these two trained NNs to determine the expected QoE to be achieved through all the potential target eNBs. The handover algorithm then triggers the handover to the target eNB for which the file download is

expected to finish successfully, and in case, there are two or more potential target eNBs, it handover to the eNB with the lowest value of the estimated file download time , hence increasing user experience and network performance as well.

6) Prediction of Fault in telecom network.

          Telecomm networks are continuously monitored and KPI has been analysed to evaluate network performance , KPIs are defined in the Network Element's maintenance manuals and vary for each NE based on its functional scope of existence, by analysing KPI and alarm mobile network operator evaluate network performance  and they will able to understand how much stable is there network is. When NE become unstable alarm will get generated and manual intervention is done to stabilize  the network element , but for the time duration  network element become unstable operator will loss revenue , as end user will not be able to use there service. With AI based techniques like distance based techniques (k-NN), Cluster analysis, Classification (SVM) techniques, or rule based techniques can be used to predict those fault which can happen in future and thus mobile network operator can take action to rectify the fault and thus both end user experience and revenue can be enhanced.

Which is not feasible from legacy network.

Conclusion

In this Article we have discuss some of the use case of Artificial intelligence techniques  like NN, linear reg., RL, SVM, cluster analysis etc. in telecom industry , which a service provider / Equipment vendor can leverage to enhance user experience of end user along with proper optimization of resource.

Since still use of AI application in telecom industry is very minimal and hence it is having a great potential to enhance the performance of network by using AI techniques in telecom domain. Which will open new opportunity for AI experts along with domain knowledge of telecom to implement these AI application in telecom industry.

I hope I have discussed some of the interesting applications of AI techniques  in the telecom sector.

If you really like this article please share with all you colleagues 

Reference :

1) Z. Zhang, K. Long, J. Wang, and F. Dressler, ''On swarm intelligence inspired self-organized networking: Its bionic mechanisms, designing principles and optimization approaches,'' IEEE Commun. Surv. Tuts., vol. 16, no. 1, pp. 513-537, First Quarter 2014.

2) Z. Zhang, W. Huangfu, K. Long, X. Zhang, X. Liu, and B. Zhong, ''On the designing principles optimization approaches of bio-inspired self-organized network: A survey,'' Sci. China Inf. Sci., vol. 56, no. 7, pp. 1-28, Jul. 2013

3) A. Imran, A. Zoha, and A. Abu-Dayya, ''Challenges in 5G: How to empower SON with big data for enabling 5G,'' IEEE Netw., vol. 28, no. 6, pp. 27-33, Nov./Dec. 2014.

4) Z. Zhao, J. Chen, and N. Crespi, ''A policy-based framework for autonomic reconfiguration management in heterogeneous networks,'' in Proc. 7th Int. Conf. Mobile Ubiquitous Multimedia, 2008, pp. 71-78.

5) S. Fan, H. Tian, and C. Sengul, ''Self-optimization of coverage and capacity based on a fuzzy neural network with cooperative reinforcement learning,'' EURASIP J. Wireless Commun. Netw., vol. 2014, no. 1,pp. 57:1-57:14, Apr. 2014.

6) A. Çalhan and C. Ceken, ''An adaptive neuro-fuzzy based vertical handoff decision algorithm for wireless heterogeneous networks,'' in Proc. IEEE 21st Int. Symp. PIMRC, Sep. 2010, pp. 2271-2276.

7) 3GPP TS 36.331, "Radio measurement collection for Minimization of Drive Tests (MDT); Overall description," version 10.4.0, Release 10.

8) "Feed-Forward Neural Networks and Multinomial Log-Linear Models," http://cran.r-project.org/web/packages/nnet/nnet.pdf.

9) S. S. Mwanje and A. Mitschele-Thiel, "Distributed cooperative Qlearning for mobility-sensitive handover optimization in LTE SON,"in Proceedings of the Computers and Communication (ISCC), IEEE Symposium on, June 2014, pp. 1-6.