Virtual RAN Control via Neural Network Resource Scheduling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing virtual Radio Access Networks (vRAN) systems face inefficiencies in resource pooling due to inelastic real-time workload scheduling, leading to over-allocation of computing resources and challenges in characterizing the relationship between radio and compute resource demands, especially with the advent of flexible 5G architectures.
Innovation Solution
A method that jointly improves compute and radio scheduling policies using contextual data such as signal quality and data demand, employing neural networks to assign resources and select modulation and coding schemes, thereby optimizing user delays and computing costs while meeting quality-of-service targets.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If RAN densification is implemented to accommodate increasing mobile service demand, then spectral efficiency is improved through spatial reuse, but capital and operating costs increase substantially
Solution Approach 1:
The patent merges multiple Radio Access Networks into a virtualized RAN infrastructure where multiple RAPs share common computing, storage, and networking resources. This consolidation allows spectral efficiency improvements from densification while reducing duplicate infrastructure costs by pooling resources across multiple access points.
Solution Approach 2:
The virtualized RAN architecture creates universal computing and storage resources that can be dynamically allocated to serve multiple RAPs simultaneously. This multi-functional infrastructure replaces dedicated hardware at each RAP with shared resources that serve the entire densified network, reducing overall capital and operating costs.
2Area of stationary object
If RAN densification is implemented, then network coverage is expanded, but management and control complexity increases due to volatile and unpredictable network load
Solution Approach 1:
The patent combines the management and control functions of multiple RAPs into a centralized virtualized management system. This consolidation simplifies control by providing unified oversight of network load across the expanded coverage area, rather than managing each RAP independently amidst volatile load conditions.
Solution Approach 2:
The virtualized RAN implementation includes mechanisms to monitor and feedback on network load conditions across the densified infrastructure. This enables dynamic resource allocation and load balancing that adapts to unpredictable traffic patterns, reducing management complexity while maintaining expanded network coverage.
3Reliability
If traditional macro-cells aggregate multiple flows, then uncertainty of individual flows is compensated, but resource allocation efficiency decreases
Solution Approach 1:
The patent implements dynamic resource allocation in the virtualized RAN that adapts to real-time flow conditions. Unlike static aggregation in traditional macro-cells, the virtualized architecture can dynamically adjust resource distribution to individual flows based on current network conditions, maintaining reliability while improving overall allocation efficiency through flexible, demand-driven resource management.
Data Source
AI summary
A radio access network includes a processing system and one or more radio access points configured to broadcast over a radio band including one or more radio channels. Contextual data is acquired which is representative of at least one of: (i) a quality of the radio band and (ii) a quantity of data demanded by user equipment in communication with the one or more radio access points over the radio band. A compute policy and a radio policy are generated based on the acquired contextual data. Data transmissions for processing are assigned to computing resources of the processing system based on the compute policy. Data are scheduled for transmission over the radio band based on the radio policy. A modulation and coding scheme for the scheduled data transmissions is selected based on the radio policy.