Huawei’s Kevin McDonnell says autonomous networks can detect issues and make decisions on their own, making them essential for future network operations.
The complexity of telecom networks makes network management increasingly complex. But autonomous networks are the next stage in ensuring that they can be managed with a more hands-off approach.
Kevin McDonnell is the senior director of AI and network autonomy at the Huawei Ireland Research Centre. His role entails bringing intelligent automation and ultimately autonomy to telecom operations.
“This means developing solutions that operators want to deploy because it solves a particular problem,” he told SiliconRepublic.com.
“Autonomous networks essentially make telecom operations smarter and more proactive. Unlike traditional networks that rely on constant human monitoring and manual configuration changes, autonomous networks can detect issues, make decisions and adapt in real time, all on their own.”
This means that when there’s heavy network traffic, an autonomous network can reroute data to prevent slowdowns and downtime. “It shifts the role of human operators from firefighting to focusing on strategic improvements.”
The tech behind the networks
Autonomous networks use a mix of different AI and machine learning models. Some are for making predictions, others handle natural language tasks.
For example, McDonnell said machine learning could be used for pattern recognition, while large language models (LLMs) or generative AI could be used to simulate potential outcomes.
They also use predictive analytics, which McDonnell said is like the network’s early warning system. “While technologies like LLMs help with understanding and generating responses, predictive models are crucial for maintaining network stability,” he said.
“They’re pretty accurate and get better over time as they learn from more data. This allows the network to proactively address problems, often fixing them before users even notice something was wrong.”
Another big focus lately is on assistive agents or ‘copilots’. These are like virtual assistants and virtual operators within a network.
“For example, a copilot might assist with tasks, while an autonomous agent can solve problems on its own. These agents can assist with tasks such as automatically rerouting traffic or managing customer requests and can trawl through data and make decisions without needing a human to step in.”
These agents would have long-term memory, unlike LLMs, which is powerful and incredibly useful. However, he warned that this is where the need for additional security comes in as the agents interact with external environments.
“This memory enables agents to learn from past experiences and adapt more effectively, but it also means handling persistent data, which comes with risks. Ensuring that these systems are secure, ethical and trustworthy is therefore part of our scope.”
Challenges
Privacy and security are absolutely critical in all aspects of technology and autonomous networks are no different, especially since they handle so much real-time data from devices all around the world.
“Autonomous networks make decisions independently, which means they often deal with sensitive information like personal locations or usage habits. To protect this data, we build in strong privacy controls and secure computing environments right from the start,” said McDonnell.
“This includes encrypting data, anonymising it and ensuring only authorised people or systems can access it. By following best practices, we aim to be transparent about how data is handled, which helps build trust with users and stakeholders.”
Autonomous networks are not without other challenges, particularly how they can handle the sheer volume and complexity of data that networks generate – and this data needs to be processed and reacted to in real time.
“Another issue is that while networks have evolved rapidly, the operations side hasn’t kept up. We still see a lot of manual processes. Knowledge is often siloed among experts or buried in documents, making it tough for systems to access and learn from it,” said McDonnell.
“Tools like intelligent assistants, or copilots, can help by gathering and centralising this knowledge. Lastly, building trust in these autonomous systems is crucial. Operators need to feel confident that the network can handle complex situations reliably.”
The future of autonomous networks
Looking ahead, McDonnell believes autonomous agents will become a standard part of all network operations, from customer service to network optimisation. He said these agents could handle complex tasks independently, moving us closer to fully autonomous networks, or what’s known as level five autonomy.
“At our Ireland Research Centre, we’re developing cutting-edge architectures, models and approaches to support these autonomous agents. Right now, we’re working towards level four autonomy in specific areas. This means the network can handle most situations on its own but might still need human oversight for complex scenarios, like handling emergency outages,” he said.
“One of the most challenging aspects is ensuring these systems are fair, unbiased and reliable. Building trust is crucial; operators need to feel confident that the agents are not only effective but also ethical.”
Don’t miss out on the knowledge you need to succeed. Sign up for the Daily Brief, Silicon Republic’s digest of need-to-know sci-tech news.