Read the related report Fusing outside plant data and AI to maximize CapEx investment for more insight into how CSPs can better extract and incorporate outside plant data into the bigger operational picture.Outside plant may be the last place you would expect to find artificial intelligence (AI), but it should be the first. The seeds of AI have been sown there, and communications service providers (CSPs) are already reaping benefits.
“There is a lack of understanding and appreciation for the things you can do with AI in outside plant today,” says Randall Frantz, Founder of RCF Consulting and former Director of Telecommunications Industry Solutions, Esri. He notes that for most CSPs the work of planning, implementing and tracking network assets is still a manual process that is labor intensive and not very accurate or consistent across organizations. But this has started to change, thanks in large part to advances in geographic information system (GIS) technology and its extensibility to other planning and engineering applications.
Benefits of GIS
Penn State’s College of Earth and Mineral Sciences defines GIS like this: “GIS can manipulate and analyze spatial datasets with the purpose of solving geographic problems. GIS analysts perform various operations on data to make it useful for solving a focused problem. These operations include clipping, re-projecting, buffering, merging, mosaicking, extracting subsets of the data, and hundreds of others.”
CSPs are using AI to make sense of all these operations and map output from them. Verizon, for example, mandates the use of GIS tools for engineering and design so that they can be stored in and accessed from a database by multiple groups, which renders computer-aided design (CAD) drawings useless for both planning and ongoing maintenance, according to Frantz.
Two other ways of using AI in outside plant planning and engineering are:
- Recognition – much of the mapping for engineering networks is now conducted by satellites and drones. Satellites gather geographic data and features, and drones are used for the same purpose, as well as for site surveys. AI is used to identify specific assets on the ground such as light poles, street signs, railways, buildings and the lineof-sight interference associated with them, plus more.
- Analysis – AI is used for analysis when multiple sources of other datasets are overlaid onto these maps to help determine where to put a cell site, route fiber or cross a river. This data includes existing infrastructure designs, planned routes, demographics, traffic patterns, real estate information (such as which building might allow cell site construction), or the general use of buildings and sites (for example, short-term, heavy but transient pedestrian traffic like at a bus stop, or long-term traffic such as a sporting event.) Some manual driving is still done by Google and others, but more targeted applications – such as Cyclomedia, which provides a cloud-based solution for street-level imagery with GIS accuracy, GPS data, high-definition and 3D imaging and other data for infrastructure, engineering, utilities, real estate and other companies – is more applicable to telecom engineers.
AI for planning
A Nokia-sponsored paper from STL Partners published in May finds that the top area for applying AI in telecom is in optimizing networks and operations, which incorporates a broad swath of functions. Similarly, a recent report by Ericsson on using AI to enhance returns on 5G network investments shows that CSPs believe AI will be especially helpful for:
- Improving network capacity planning and management
- Reducing time for network planning
- Improving infrastructure modelling to reduce CapEx
TM Forum’s research reinforces these findings. In 2018 we surveyed CSPs to find out how they were using AI in operations and how they expected to use it within two years. Nearly three-quarters said they were already using machine learning for network planning, optimization and management.
The graphic below shows the breakdown by type of AI. We often lump tools such as AI, machine learning and big data analytics together under the umbrella term AI, but when it comes to analyzing deployment it is necessary to break them out. The graphics opposite show that CSPs are clearly making a distinction between machine learning and deep learning AI platforms.
Basic machine learning techniques are more mature than deep learning, which few CSPs said they are currently implementing. However, operators plan to expand their use of deep learning, with nearly a third saying they will use it by the end of next year.
