How can AI agents ensure a self-driving and more sustainable future for networks?

Energy efficiency is a fundamental expectation for communication service providers. Advanced AI algorithms that learn how to control our networks gives us new opportunities for optimisation and reduced energy consumption in our networks.

By testing and applying cutting-edge autonomous AI control solutions, we can move towards enhancing energy efficiency without compromising on the quality of connectivity in mobile networks. This work extends on earlier Telenor "Green Radio" initiatives where more basic AI methods were used to optimise the network. The recent advancements in AI allows us to develop a far more complex control system - a technology similar to that used in self-driving cars.

The team has been able to create a simulator that acts as learning environment for training an AI agent. During millions of trial-and-error attempts in this simulator, the agent learns how to minimize power consumption without sacrificing network performance. Currently, a generalized AI agent has been developed and trained. It is now being adapted to the "real world" to apply its learning to control live network sites.

Preliminary results from the work indicate 3% incremental energy saving potential with AI control, and 17% less violations of keeping 20 Mbps throughput constraints. The work is still to be verified in a real world environment.

By leveraging advanced AI control solutions and moving towards enhanced energy efficiencies in the communication networks we hope to take a step towards a more sustainable future for the networks and society.