Scribd
Upload
145 views

Non Real-Time RAN Intelligent Controller (RIC non-RT)

The Non Real-Time RAN Intelligent Controller (RIC non-RT) project aims to support non-real-time radio resource management and optimization in RAN. Key functions include network analytics, model training, and mapping non-real-time policies and AI/ML models to guide the RIC near-RT for localized decisions. The project is part of the O-RAN Software Community and uses ONAP with interfaces to wireless network elements defined in O-RAN specifications. Initial scope includes enhancements to support use cases without code impacts to ONAP Release 4.

Uploaded by

ESUMIVI
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
145 views

Non Real-Time RAN Intelligent Controller (RIC non-RT)

The Non Real-Time RAN Intelligent Controller (RIC non-RT) project aims to support non-real-time radio resource management and optimization in RAN. Key functions include network analytics, model training, and mapping non-real-time policies and AI/ML models to guide the RIC near-RT for localized decisions. The project is part of the O-RAN Software Community and uses ONAP with interfaces to wireless network elements defined in O-RAN specifications. Initial scope includes enhancements to support use cases without code impacts to ONAP Release 4.

Uploaded by

ESUMIVI
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 7

O-RAN Software Community

PROJECT PROPOSAL
For inclusion from Release A

Non Real-Time RAN Intelligent Controller (RIC non-RT)

Project Name:
• Proposed name for O-RAN SC project:
Non Real-Time RAN Intelligent Controller (RIC non-RT)
• Proposed name for the repository:
non-rt-ric

Project description:
The primary goal of RIC non-RT is to support non-real-time intelligent radio resource
management, higher layer procedure optimization, policy optimization in RAN, and
providing guidance, parameters, policies and AI/ML models to support the operation of
RIC near-RT to achieve higher-level non-real-time objectives. RIC non-RT functions
include service and policy management, RAN analytics and model-training for the RIC
near-RT functionality.
With the amount of L1/L2/L3 data collected from eNB/gNB (including CU/DU), useful
data features and models can be learned or abstracted to empower intelligent
management and control in RAN. For example, network spatial-temporal traffic patterns,
user mobility patterns, service type/patterns along with the corresponding prediction
models, network quality of service (QoS) prediction patterns, massive MIMO
parameters configuration, and more can be reused, abstracted or learned. This
information can then be combined with additional network-wide context and policies to
drive fine-grained near-real-time network radio resource management in the RIC near-
RT and non-real-time optimization within RIC non-RT.
Analytics, intents and policies defined for the RIC non-RT in many cases will need to be
mapped to sub-policies, actions, parameter changes, analytics/learning tasks, etc., to
steer the operation of the RAN via the RIC near-RT. Such changes will enhance the
local analytics/learning and decision processes in the RIC near-RT to be enriched by
higher-level context and goals. For example, localized near-real-time decisions about
load-balancing, ID allocations, handover decisions, resource allocations, etc., can
greatly benefit from adjustment and guidance about the high-level network-wide intents
of the network operators. The core algorithms executing in or above the of RIC non-RT
will be developed and owned by operators, some aspects of which may be proprietary.
This will provide the capability to modify the RAN behaviors by deployment of different
policies and models optimized to individual operator intents and objectives. The policies
passed from RIC non-RT to RIC near-RT are created, modified and deleted by the RIC
non-RT, guided by context information or policy state information received over A1 or
O1, e.g. from the RAN indicating intent fulfillment (O1).
The A1/O1 interfaces together will support communication & information exchange
between the RIC non-RT and the RIC near-RT, while also supporting the exchange of
FCAP information to the Orchestration and Automation layer above. The use of the
A1/O1 interface must be coordinated in a holistic manner.
Key functions that the A1/O1 interfaces are expected to provide include:
• Network & UE-level information/context exposure from network elements to RIC
non-RT to support various requirements such as network management, online
learning and offline training of AI/ML models and driving non real-time
optimization into the network.
• Support for policy-based guidance of RIC near-RT functions/use-cases,
deploying/updating AI/ML models into RIC near-RT, and feedback mechanisms
from RIC near-RT to ensure that operator’s intents and objective are met.

Scope – Release A:
The following features are in scope for the RIC non-RT project within O-RAN SC
release A:
• Enhancements to support O-RAN SC release A use cases that can be delivered
“on top of” ONAP release 4 Dublin (in other words, without Dublin code impacts):
o Yang models based on O-RAN IM/UML or YANG files imported from O-
RAN WGs
o Directed graphs (DG) for SDN-R / CCSDK
o New Adapters needed to support use cases (details to be determined
during planning phase)
o Netconf, Ansible, and Chef interfaces
o Initial ML model catalogue access for model training performed for RIC
non-RT (details to be determined during planning phase based on
identified use cases)
o Non real-time automated control via CM changes in network functions
based on an initial O-RAN Yang model as an output of the analytics
performed for the RIC non-RT (ML-assisted or otherwise, exact use case
to be determined during planning phase)
• Support for third party controllers
o Adapter to allow DG to connect to NetConf devices

• High availability (local)

• The following features may be defined for this project:


o Configuration versioning: ability to roll back the configuration
o Support for third party controllers
▪ Adapter layer to interface with downstream controllers
o Support for geographically distributed network resources

Scope / Ambition – 2020+


The RIC non-RT will form a significant part of the Orchestration and Automation (O&A)
layer of O-RAN, situated on top of the A1 interface. In addition to the RIC non-RT
activities already mentioned the O&A layer together with the RIC non-RT must also
handle configuration and deployment FCAP activities (e.g. data collection for
information to be received over O1 interface). There must exist functions that handle
AI/ML workflow, e.g. training and update of ML Models, both inside the O&A layer and
for deployment to RIC near-RT. There must also be access to other non-RAN related
data to enhance the RAN operation and optimization functions, again within the O&A
layer and the RIC near-RT.
The RIC non-RT must support these activities in the O&A layer. Some longer-term
questions that must be addressed include:
• How are non real-time control loops designed, modelled, packaged, deployed,
monitored, and adapted after deployment?
• How are the high-level operator goals/intents mapped to actionable/deployable
policies for the RIC near-RT?
• How are the monitoring activities to monitor intent satisfaction derived from
intent-based policies, then deployed and managed?
• How are ML/analytics models in the O&A layer and RIC near-RT coordinated,
dynamically updated and tuned in a coordinated manner?
• Automated decomposition of intents and ML/analytics functions requires accurate
comprehensive details of underlying target infrastructure and capabilities. How
will these models be specified, populated and updated in a coordinated manner
to achieve this decomposition?
• How are conflicting models or intents handled in a coordinated manner?
• How can proprietary/black-box models/policies/services/loops be handled in a
manner that enough detail is available to support their coordinated operation
while protecting their internal operation?
Architecture Alignment:
How does this project fit into the rest of the O-RAN Architecture?
This project uses the O-RAN A1/O1 interface specifications to integrate wireless xNFs
with RIC non-RT which is a component within the Orchestration & Automation layer
depicted within the O-RAN architecture diagram below:
RIC non-RT is built on ONAP and intended to be used within ONAP; however, there are
other options within the Orchestration & Automation layer.

How does this align with external standards/specifications?


TOSCA, NETCONF, YANG, and the O-RAN Information Model (OIM) for control of
multi-vendor wireless network elements through open management interfaces, as
defined in O-RAN WG interface specifications.
Are there dependencies with other open source subprojects?
• Acumos
• Akraino
• Open RAN Alliance
• Open Daylight
• ONAP
• Open Networking Foundation
• OPNFV

Resources:
Project Technical Lead (PTL):
John Keeney (Ericsson) [email protected]
Names, gerrit IDs, and company affiliations of the committers:
Martin Skorupski (highstreet), [email protected]
George Clapp (AT&T), [email protected]
Dongho Kim (AT&T), [email protected]
Herbert Eiselt (highstreet), [email protected]
Raymond (Zach) Lovell (CommScope), [email protected]
<completion of roster in progress>

Names and affiliations of any other contributors (in alphabetical order):


Dhruv Gupta (AT&T), [email protected]
Thorsten Heinze (Telefonica), [email protected]
Petr Jurcik (DTAG), [email protected]
Lyndon Ong (Ciena), [email protected]
Morgan Richomme (Orange), [email protected]
Slawek Stawiarski (AT&T), [email protected]
Tracy Van Brakle (AT&T), [email protected]
5X Ericsson Developers
<completion of roster in progress>

We propose weekly open meetings chaired by the PTL (Current proposal: Mondays
16:00-17:00 UTC).
Project requirements
The project requests the following facilities form Facilities from LF/ORAN SC:
• Mailing list, or hashtag: https://lists.o-ran-sc.org/g/main
• Confluence site: https://wiki.o-ran-sc.org/display/RICNR (TBC)
• JIRA: https://jira.o-ran-sc.org/
• Gerrit repo: https://gerrit.o-ran-sc.org/r/admin/repos
• Slack Channel
• Zoom account

Lifecycle State: incubation

You might also like