Description
The objective of this activity is to develop a machine learning (ML) engine that optimises handover between two different 5G networks, where at least one of them is a satellite-based non-terrestrial network (NTN). The activity will also provide the testbed to assess the handover key performance indicators in a laboratory environment.Targeted Improvements:Identify and avoid higher risk handovers improving thus the handover success rate to reach at least 95%.Description:The coverage extension promised by the integrationof terrestrial and non-terrestrial 5G networks comes with technical challenges. Ubiquitous network access requires 5G user equipment capable to connect to both terrestrial and non-terrestrial networks directly. Handheld devices with omni-directional antennas operating in lower frequency bands is expected to support dual connectivity. Such devices will need to handover 5G connections between terrestrial and non-terrestrial networks. The number of handovers between gNBs and the success rate is important for the network as it influences the service unavailability. In low earth orbit (LEO) NTN deployments with satellites bearing regenerative payloads, thesource gNBs can be implemented on-board while the target gNBs on ground. To initiate the handover process, the source gNB will senda request to the target gNB. In practise, a handover can be triggered by signal strength measurements, location information and other parameters such as Timing Advance (TA) values or doppler values exceeding a predefined threshold. Moreover, the handover rate maydepend on the inclination orbit of the satellites, the antenna features, but also on the connection interval. Often, the network needs to take handover decisions while dealing with partial information, uncertainty, time constraints, and rapidly changing communication environment. The targeted machine learning solution shall assist the 5G NTN in developing the required cognition to improve thehandover success rate to over 95%, decreasing thus the handover signalling load and the service interruption time to a minimum. This activity shall develop and test a machine learning (ML) engine capable to optimise the decision to handover connections between two different 5G networks, where at least one of them is a satellite-based non-terrestrial network (NTN). The associated development shall be of a modular architecture to allow future enhancement and shall be independent of the selected hardware and software implementation platform. The activity shall provide the testbed and means to validate the ML-based handover engine and verify its performance in a laboratory environment using commercially available computer platforms.