APPROFONDIMENTO

Enabling software intelligence all over the wireless access: the O-RAN initiative

Enabling software intelligence all over the wireless access: the O-RAN initiative
 

Open RAN Alliance (http://www.o-ran.org) is an industry initiative resulting from the merging between xRAN Forum with the C-RAN Alliance, with the objective to form a world-wide, ‘carrier-led’ effort to push more openness into the radio access network of the next-generation wireless systems. It is supported by AT&T, China Mobile, Deutsche Telekom, NTT DOCOMO, Orange, Verizon and TIM and other telco operators.
The openness of network architecture and interfaces is carried out in two main tracks:

  • Defining implementation requirements by means of “profiles” defined on top of Specifications provided by the relevant Standard Development Organizations, like 3GPP, so that multi-vendor interoperability of RAN functions is ensured. This aspect involves both functional interfaces and management interfaces
  • Introducing new functions and interfaces to centralize the key radio resource management functions in a RIC (RAN Intelligent Controller) interworking with the existing RAN functions (e.g. gNB-CU and DU) where customized algorithms, empowered by the adoption of Artificial Intelligence / Machine Learning techniques, can be designed and implemented as Cloud Native Applications.

Near Real Time RIC exposes RAN capability and resources towards the Management and Orchestration layer over A1, an “intent-based” interface used to instruct RAN with performance targets and Radio Resource Management policies. The new A1 interface will complement Network Management and Orchestration O1 interface providing fault-management, configuration, accounting, performance, and security (see figure).

 

O-RAN Architecture: Example Use Case “3D MIMO Beamforming Optimization”

The enforcement of RAN policies is then performed by coordinating resources and algorithms in the RAN functions as defined in the 3GPP RAN Architecture, i.e. CU-CP, CU-UP and DU. In addition to 3GPP architecture, O-RAN architecture introduced a further functional split of the Layer 1 protocol stack in the DU over an Open Fronthaul interface, resulting in the separate O-DU and O-RU functions.
The modular, open, intelligent, efficient, and agile disaggregated radio access network resulting from the O-RAN reference architecture is expected to facilitate the exploitation of Hardware/ Software separation and the adoption of open source solutions.
With this aim, the “O-RAN Software Community (SC)” (www.o-ran-sc.org) was created as a collaboration between the O-RAN Alliance and Linux Foundation with the mission to support the creation of software for the RAN (Radio Access Network). The O-RAN SC plans to leverage other LF network projects, while addressing the challenges in performance, scale, and 3GPP alignment. Coordination and synergic effort is also expected towards ONAP and 3GPP with particular focus on Network Management and Orchestration architectures and platforms.
The use case “3D MIMO Beamforming Optimization”, proposed in O-RAN by TIM, Orange and CMCC is an example of how 5G M-MIMO capability can be exploited to adapt the network coverage and capacity based on service requirements and traffic distribution (see figure):

  • a dedicated function (Non Real Time RIC) in the management layer retrieves necessary configurations, performance indicators, measurement reports and other data from Configuration and Performance Management for the purpose of constructing/training relevant AI/ML models. Moreover other information from the application layer (e.g. GPS location of users) and mobility pattern predictions can be used to enrich the model;
  • the trained AI/ML model is transferred to the near-Real Time RIC and may be used to infer the user distribution of multiple cells, and predict the optimal configuration of radio resources and M-MIMO parameters for each cell according to a global optimization objective designed by the operator;
  • optimal beam pattern configuration and/or policy are then applied by changing the configuration in the affected CU/DU/RU;
  • the fulfillment of the network performance is checked against the new targets and the adopted configurations. Periodic model retraining and policy adjustment is then performed to cope with the dynamic nature of the traffic and the user distribution.
 

Torna all'articolo