How AI helps to control the outside plant of a PON network
In the first blog in this series, we gave an overview of how Artificial Intelligence (AI) and Machine Learning (ML) are transforming fiber broadband network operations, creating networks that can sense, think, and act. In this and subsequent blogs, we’ll dive deeper into some of the most compelling uses cases: data analysis, autonomous operations, digital twins, and beginning here with assistance to humans.
Broadband operators struggle to keep inventory systems up to date for the passive plant in the field at various levels: street cabinets, distribution points, splitters, etc. When a new subscriber is added – it is essential to know how much free capacity is still available and determine if any extensions are required prior to making an appointment with the end user. Inaccurate data can lead to additional truck-rolls, an increased number of repeat visits to connect subscribers to the fiber plant, or performing extensions that may not be required.
Even more, in open access networks where multiple retailers connect subscribers at the same flexibility point of the infrastructure provider – typically with manual reporting between the various parties - the risk on misalignment between multiple parties and therefore inaccurate deployments are higher. At the street cabinet – it may be difficult to know which end user is connected to what ports – resulting into ‘spaghetti’ troubleshooting and slow problem resolution. This may result into the need for a full audit of the outside plant to determine what ports are in use or free, and what subscribers are connected to which ports.
In general the installation quality of the passive components is difficult to monitor, and acceptance of installations is largely done on partial audits, likely performed after the technician has left the site. Low quality installations may lead into performance degradation or failures over time..
Since passive equipment by definition is not monitored and does not generate alarms, there are no straightforward ways to solve this. This is where AI – and AI computer vison in particular - comes in.
Computer vision is a branch of artificial intelligence focused on enabling machines to analyze, process, interpret, and understand visual data from images and video. Computer vision comes into its own when field technicians are equipped with smartphones with high-quality cameras. They can capture equipment before and after interventions , with the images then sent for analysis using object detection, image recognition, optical character recognition (OCR), barcode recognition, and data matrix code recognition. This can aid in a wide variety of activities.
How can you start with AI-assisted human interventions? Don’t build Rome in a day. Start with the use cases that provide the highest benefits for your operations.
-
one priority may be to control and manage the capacity of the passive plant, and in particular the filling rate of distribution points, splitters and street cabinets. .. A picture taken by the field technician during the initial installation of the OSP is compared with follow-up pictures taken while subscribers are connected. At any point in time the AI model can detect which ports are in use, which ports are still free, and therefore how many subscribers can still be connected. Any deviation versus the inventory system can be spotted and analyzed in real-time by the technician in the field.
-
Next, frequently occurring installation errors can be detected, in particular errors that may cause degeneration of quality over the lifetime of the component, such as missing dust caps impacting the quality of the fiber connectors over time, missing screws, fiber bends, etc. Performing automated checks will lead to better quality of the outside plant and motivate field technicians to adhere to installation procedures – leading to continuous improvement. Field technicians can be guided and trained in the field by the AI assistant.
-
Another relevant use case can be to identify which subscribers are connected to specific ports at the distribution point and at the street cabinet. Patch cord labels or tags can be read out and matched with a specific subscriber. This information is essential for trouble shooting and analysis - remote or in the field.
The good news with AI is that there is a means to start small and evolve. Not a revolution, but an evolution. Use AI where it makes business sense – and build from there.
Watch out for the next blog in this series and, in the meantime, download this white paper to learn more. Also, why not check out the Nokia Broadband Easy Connect platform – a digital platform to simplify and streamline the fiber home connect program, powered by AI.