The main objective of this proposal is to implement an AI/ML based system to manage the dynamic allocation of resources to different slices in the Core of a 5G network. This system will take actions in two different aspects: it will be the responsible of scaling up and down different User Plane Functions (UPFs) associated to a specific slice, and it will be the responsible of reallocating the physical resources of the infrastructure to the virtual slices depending on the level of saturation of the 5G network or a specific slice. The proposed system will integrate two different types of ML algorithms: 1) Incremental Learning (IL) based algorithm for streaming time series forecasting regarding both users and state of the network; and 2) a Reinforcement Learning (RL) agent to automatically and optimally manage the 5G network resources, considering current and future states. To this end, the outputs of the IL model will feed the RL agent.
