Wednesday, June 13, 2018

Wireless Network Balancing - Impact of Artificial Intelligence & Machine Learning


Mobile networks today are highly complex in nature. With evolution of networks from 3G to 4G and explosive subscriber growth, the amount of data flowing through the networks have increased manifold. Along the way, networks have become denser with introduction of new solutions such as femtocells and picocells (personal base stations) that boost mobile network coverage and provide additional capacity. This is where the concept of Self Organizing Networks or as it generally referred to as SON was developed. Old ways of managing the network in terms of service availability or network capacity planning are not scalable. The mobile first world needs an innovative approach to the way modern day mobile networks are managed and the way network coverage and capacity is planned. Similarly, operators need to find ways to reduce CAPEX and OPEX without sacrificing on network quality and maintenance needs.

SON is a concept (set of highly complex algorithms) where automated processes are used to monitor the network and measure its performance and network analytics is used to gauge the feedback for making critical decisions that help manage the complex network and reduce the costs.  There are 3 main areas over which mobile networks use SON algorithms for self-optimizing and balancing:

Self Configuration
Self-configuration allows wireless base stations or access points to be plug and play. They need as little manual intervention as possible.

Self-Optimization
Radio resource is often the bottleneck when it comes to mobile networks. With SON, the radio resources are managed efficiently and intelligently thereby providing optimization of coverage, capacity and interference.

With SON, networks can perform mobility load balancing where by cells which are heavily loaded can transfer load to other cells which are lightly loaded thereby achieving the optimal balance. This load balancing technique can be used even with different radio technologies e.g. between 4G and 3G networks and also during handovers (Handover take place for e.g. when the call is active and you are driving from one place to another without dropping the call.)

Self-Healing
Any complex system experiences intermittent failure or faults/errors. This can lead to service interruption or poor service quality. Self-Healing involves automatic correction of network parameters and removal of failures to mask the effect of the fault or failure. This self-healing capability provides the necessary stability and reliability for mobile networks.

With advent of Machine learning and Artificial intelligence, SON can be done at much larger scale and with more accuracy. Imagine a scenario where a region is facing poor signal coverage or low throughput numbers. While such a issue would have needed much intervention from engineering, going forward smart algorithms can isolate the root cause of such issue and lets say if the cause is some configuration mismatch or routing issue, they can fix the issue without manual intervention. 

AI & ML will be more important as today's wireless networks evolve to 5G. 5G will result in network densification and there will be tons of wireless nodes deployed (Macro cells, femtocells etc) and service automation will be key in scaling of next generation networks and their day to day operation. 

No comments: