The great value of automated site configuration

by | Feb 18, 2018 | Use Cases

Mobile networks develop every day. The task to generate and change site configuration data is time demanding and repetitive. The larger and more granular the network gets, the more unmanageable it becomes. Site configuration tasks are therefore perfect for automation. In this blog, we look at the aggregated effect of three automation use cases: ‘Physical Cell Identity conflict resolver’, ‘Automatic Neighbour Relation optimisation’ and ‘Scrambling Code optimisation’. The result? The dropped call rate improved, the number of customer complaints decreased and the reduction in manual workload led to cost savings. We’ve done it. Now it’s your turn.


Three orchestrated site configuration use cases 

The only thing that is constant is change. This is also true for radio networks. Like for most mobile operators, our networks in Finland and Estonia are developing every day. Sites, cells and frequencies are added. Every change has ripple effects on the surrounding cells. The task to generate and change site configuration data is time demanding and repetitive – and the larger and more granular the network gets, the more unmanageable it becomes.

“Site configuration tasks are therefore perfect for automation”, explains Dr. Jarno Niemelä, the technical lead of the automation. “It replaces a work element nobody really loves anyhow, saving time and money and improving the customer experience”.

To show that also such use cases can make a big difference we are here looking at three use cases together:

  • Physical Cell Identity conflict resolver
  • Automatic Neighbour Relation optimisation
  • Scrambling Code optimisation

Physical Cell Identity conflict resolver

Every cell in a 4G LTE network is assigned an identity – called the Physical Cell Identity or PCI. There are however not enough identities to cover all cells in a network – the maximum number of PCIs is 504. The cell identities therefore have to be reused. This means that there is a constant risk for conflicts if two cells too close to each other have been assigned the same PCI – conflicts that lead to dropped calls, connection issues and dissatisfied customers.

Networks aren’t static. There’s always rollout going on – and the growth in data traffic and the introduction of new technology such as 5G will mean more, not less, rollout. On top of this, there’s also changes incurred by e.g. Automatic Neighbour Relation optimisation that we introduce in the next section.

We used our software development kit (SDK) to create an algorithm that automatically detects PCI conflicts – and resolves them. The principle behind the algorithm is to detect all PCI collisions on a daily basis, randomly select a new PCI value where there has been a conflict (excluding the PCI value of the neighbours, of course) and then iterate the algorithm a number of times in order to avoid neighbour-neighbour collisions. The PCI algorithm runs in a closed loop so it doesn’t need any human interaction to calculate and optimise new parameters.

The implementation of the algorithm had an immediate impact on the performance of our network in Finland.

When the algorithm was first launched in 2015, it detected thousands of PCI conflicts – both collisions and confusions. Our network was actually well planned – manually – by experienced engineers, yet the automatic algorithm found issues that the human eye hasn’t been able to spot. But as the algorithm started to resolve these, there was an immediate drop in the number of conflicts detected. As it checks the entire network every day, it didn’t take long until the number of PCI conflicts were down to one tenth of what it was before.

‘This is just not possible to achieve with manual labour.’


The effect on the handover success ratio was incredible. The already strong KPI – around 98% – was improved to a level of about 99% by implementation of our closed-loop algorithm. This is just not possible to achieve with manual labour. And when handovers between cells are successful, it pays off in customer experience. The dropped call rate decreased significantly – from a level above 0.2% down to a level of about 0.05%. In 2017, the dropped call rate in our network has decreased even further and is now below 0.05%. Any network planner knows that this is a world-class value.

“In two years we cut the dropped call rate to less than one fourth of what it was. The PCI conflict resolver had a major impact and we are so proud of the results”, says Dr. Jarno Niemelä.

Automatic Neighbour Relation optimisation

The standardisation body 3GPP has specified something called ‘Automatic Neighbour Relation’ or ANR. It exists for both 3G and 4G LTE networks. It offers help to the network planners of operators as it automatically detects and proposes which cells should be defined as neighbours – and thereby have handover relations.

The problem is that ANR creates a lot of neighbours – too many. The terminals of the customers will spend much effort evaluating neighbouring cells that are seldom used. Instead they should keep close track of the most used and appropriate neighbours.

To address this, we created use cases that enhance ANR for 4G LTE as well as 3G. Our 4G LTE algorithm automatically limits the number of neighbours that ANR produces. It has improved the handover success rate and the user satisfaction.
Our 3G algorithm similarly limits the number of neighbours by removing unnecessary neighbours based on utilisation. We removed more than 120.000 unused adjacencies in our network in Finland, leaving room for new cells to be added to the network in the future. But it also saved us from dedicating a network optimiser full-time to the task.
We also automatically optimise the adjacency list (the SIB) which has resulted in a significant improvement in the setup success rates. It also saves us manpower – at least 1.5 full time employees a year.

Finally it includes our automatic intra- and inter-system adjacency creation. When we first took it into use we tested it on our 252 worst performing 3G cells. After one week of automatic optimisation, the number of dropped calls had been reduced by 46% on these cells. Again, it also saves manpower – in our case two full-time resources.

Overall, the results of our ANR optimisation efforts[1] in 3G are striking

[1] In combination with the previously described PCI conflict resolver and the Scrambling Code optimisation

During 2016, our minutes per drop – the time a customer averagely has to talk before experiencing a dropped call nearly doubled –from about 1300 to about 2500. For the average customer, that’s an additional 20 hours of talk time before a call drops.

Scrambling Code optimisation

Finally the last of the three use cases we chose to group together in this blog. In 3G networks, scrambling codes are used to distinguish different cells in downlink – and different end-user terminals in the uplink. Similarly as with the previous described PCI, the number of scrambling codes is limited and the codes therefore need to be reused in the network. As with the PCI, this occasionally leads to conflicts.

Our algorithm to optimise the scrambling codes is completely automated and works in synch with our just described Automatic Neighbour Relation optimisation for 3G. The algorithm detects the conflicts up to third tier. For each scrambling code the shortest re-use distance is calculated and as a resolution the new scrambling code is randomly selected among those with the longest.

As a result, the number of dropped calls has been reduced and the success ratio improved.

“Before we started, one fifth of the workload in our network optimisation team was related to scrambling code optimisation”, explains Dr. Jarno Niemelä. “This activity is now fully automated, saving us time and money”.


These three use case – Physical Cell Identity conflict resolver, Automatic Neighbour Relation optimisation and Scrambling Code optimisation – resulted in the following longer-term benefits:

  • Improvement in dropped call rate
  • 20% reduction in network-related customer complains[1]
  • No need to optimise Physical Cell Identities, neighbours and scrambling codes due to rollout or other changes to the network
  • 12% savings in optimisation and rollout OPEX as a result of the associated reduction in manual workload

We’ve done it. Now it’s your turn.


In combination with the LTE load balancing use case

Pin It on Pinterest