Deliverable 2.7 Initial AI-Driven Edge Computing and Precise Localization Demonstrator

The purpose of this deliverable is to give an overview of how the indoor positioning solution will be provided for the relevant use-cases (UCs) within 5G-TIMBER project.

At first a brief overview of the goals and intermediate results of other similar, 5G-related projects is provided for comparison. 

The deliverable brings out the issue about unavailability of hardware, like radio network elements and user equipment (UE), supporting 5G based positioning. Alternative approaches are explored and the most suitable ones are recommended as temporary solutions. Those temporary measures are to be used until 5G-based positioning becomes feasible. Hopefully this happens during the timeline of the ongoing project. 

Topics of edge computing, machine learning and non-line of sight (NLOS) detection are covered briefly as they all are related to either enabling positioning or improving its accuracy, availability, and update rate. 

To test the use cases demanding accurate indoor positioning during the project lifetime, an ultra-wideband (UWB) based solution is the most viable option and should be used. As one of the goals is to provide the 5G-based solutions for all manufacturing needs, including indoor positioning, then work on enabling 5G based positioning is also continued as well. 

Radio Access Network (RAN) provided at the first phase of the project by the partners is not suitable for 5G RAN dependant positioning due to this the involvement of mobile network provider Elisa with its own RAN equipment is necessary. With such involvement, a Downlink Angle of Departure (DL-AOD) method is the most suitable candidate for 5G RAN dependent positioning solution at the beginning of the project. Uplink Angle of Arrival (UL-AOA) method should be implemented in the later phase if the necessary equipment (either from the project partners or Elisa), becomes available. 

Overall, both time- and angle-based methods will be tested. Precise positioning data will be offered at RAN intelligent controller (RIC) to exploit positioning for various purposes (e.g., predictive handovers, interference management etc). Necessary computational resources for implementation of positioning solutions, including capability to support Machine Learning (ML) solutions must be available at the network edge. Demand arises as the positioning results are often time-critical.