As CSPs begin to deploy cloud-native systems and other advanced technologies, their networks are becoming more complex, putting operation and maintenance (O&M) personnel under pressure. In the radio network, for example, configuring base station parameters can now be a lengthy procedure making it difficult for CSPs to keep pace with rapid changes in the network environment, which can quickly result in deteriorating end user experiences.
This Catalyst is looking to use generative artificial intelligence (genAI) to address these challenges in a number of ways, while applying zero-trust and zero-risk principles in the network management resource layer. For example, the project is developing a series of digital assistant and digital expert solutions to ensure high stability in the core network and greatly improve routine O&M efficiency.
For the radio network, the Catalyst is developing an assurance system which will use data compression technology to quickly identify network status changes and accumulate core data. This mechanism will be supplemented by a decision-making system, based on deep reinforcement learning and large AI models, which will be able to rapidly optimize the network to meet multiple objectives and perform closed-loop management. The Catalyst will also employ machine learning to automatically expand parameter ranges and optimization objectives, while absorbing expert optimization experience to improve the model’s performance.
For the bearer network, the Catalyst will employ natural language processing technology to automatically identify customer intentions, select APIs and set parameters. The proposed solution will be able to query fault-related information through a mobile application running on end users’ cellphones, greatly reducing the time it takes to obtain fault information and the mean time to repair. The Catalyst team plans to measure the project’s feasibility and effectiveness by tracking the work order automation rate, the work order processing duration, the fault handling duration and other indicators. The goal is to automate 80% of service fault diagnoses, while enabling real-time responses to fault information queries, leading to a 60% improvement in O&M efficiency.