AI Innovation Supports Rural and Remote Internet Connectivity

Consumer internet consumption continues to soar, as more people worldwide work, attend school and connect with loved ones online. In rural and remote areas, beyond the reach of wireline connections, satellite makes this essential internet access possible. Increasingly, machine learning (ML) and artificial intelligence (AI) play an important role in optimizing satellite internet connections for Remote Internet Connectivity and providing a more reliable, higher quality end-user experience.

With more than a million households using satellite internet in the Americas alone, satellite internet providers continuously amass enormous amounts of data. Applying ML and AI to these data lakes can turn insights into intelligent actions that ensure quality of service. For instance, one challenge of satellite internet service is potential signal interference due to rain or other inclement weather. Companies like Hughes Network Systems, LLC (HUGHES) can employ ML and AI to intelligently identify weather patterns and forecast events that could disrupt service, allowing the system to switch between ground gateways proactively before a rain fade event and maintain seamless service for users. Algorithms include empirical, statistical and fade-slope models that consider rain rate, related atmospheric condition and long-term data on signal attenuation. In combination, ML and AI effectively predict near-future rain events to mitigate impact. In order to more accurately forecast further into the future, machine learning specialists are conducting promising research on a deep learning (DL)-based architecture to forecast future rain fade using satellite and radar imagery data and power measurements of satellite links.

In addition to preventing signal antennation and service disruption, ML and AI also contribute to quality of service via network triage – identifying types of data traffic and automatically deciding on how to prioritize that data. For example, the demand for streaming video content has never been higher, driven by the explosion of entertainment options like Netflix and other streaming services and the uptick in work-at-home video conferences. Using ML and AI, a satellite system can recognize data traffic that is streaming video and prioritize it accordingly, so users experience less buffering or delay. The next frontier in this kind of AI application is determining how to classify encrypted traffic accurately, maximizing the effectiveness of network triage to guarantee the best user experience with Remote Internet Connectivity.

Quality of experience estimation is another area of research and development. For instance, if someone is watching a video and there’s an issue between the server and the client and it keeps buffering, it’s important to identify and rectify the issue. But identifying the cause – such as congestion on the satellite network or an issue with the Local Area Network (LAN) – is a crucial first step. Engineers are advancing AI solutions that can pinpoint the location of the issue, then automatically take proactive steps to promote an acceptable user experience.

Another, even more challenging aspect of quality control is applying ML to physical layer technologies. For example, today’s internet experience involves lots of components like amplifiers, equalizers and other audio devices that are tough to monitor mathematically. Ultimately the goal is to model with ML so the network can compensate for imperfections. Any advances in the areas of classifying encrypted traffic and applying ML to the physical layer will not only impact satellite internet users, but relate to the growing Internet of Things (IoT). Advances made by satellite operators will benefit IoT devices as well.

In another practical application of AI and ML, satellite operators are utilizing the technology to improve customer service on the operational side of the business. Massive amounts of data result in new levels of customer insights – insights that inform each stage of the internet customer journey. For instance, if a satellite internet user is renewing a contract, customer service agents can tailor service plans based on learnings from each user’s unique usage pattern. Hughes has applied ML to equipment installation, as well, analyzing data to determine the ideal configuration of a satellite dish installation. Now, AI analyzes customer-submitted photos to identify causes and provide solutions to common installation-related problems like signal quality, malfunctioning network devices or improperly placed modems.

Perhaps the most impactful ML and AI customer service application is the ability to proactively triage network issues to mitigate customer complaints. Identifying a pattern between an influx of service calls on the business side and any network anomalies, AI prompts the Remote Internet Connectivity service provider to address issues before more disruptions occur and more service calls come in. As a result of deploying AI for network triage, the Hughes team reduced its issue detection time by 60% and repair time by nearly 50%.

AI and ML innovations in telecommunications are not limited to just enterprise business networks. They are an essential to providing people all over the world with affordable, reliable internet connectivity. Providers that prioritize innovation in ML and AI deliver a better overall customer experience and are pioneering advances that have broad applications that will continue to transform our global networks.