Network Slice Resource Allocation via Reinforcement Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current communication systems, particularly in 5G networks, face challenges in efficiently allocating network resources to meet user Quality of Experience (QoE) requirements, leading to over-provisioning and wastage of resources, as existing QoS provisioning methods are not adaptive enough to accommodate diverse QoS needs of various applications and services.
Innovation Solution
A method that involves obtaining and providing information on the mapping between network resource allocations and user QoE, as well as QoS tolerance, to a controller for allocating network resources, using a machine-learning based multi-agent architecture and flexible QoS architecture that supports non-standard applications, allowing for optimal resource allocation and QoS parameter negotiation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional QoS provisioning methods are used to allocate network resources, then network resources are allocated to meet user QoE requirements, but this leads to over-provisioning and wastage of resources
Solution Approach 1:
The patent implements dynamic QoS parameter adjustment based on real-time network conditions and user QoE feedback. The system continuously monitors network state and user experience, then adaptively modifies QoS parameters rather than using static provisioning, enabling the network to respond flexibly to changing demands and avoid over-provisioning
Solution Approach 2:
The system employs feedback mechanisms where user QoE measurements are collected and used to adjust QoS parameter settings. This closed-loop control allows the network to learn from actual user experience and optimize resource allocation accordingly, preventing both over-provisioning and under-provisioning of network resources
2Adaptability or versatility
If existing QoS provisioning methods are used, then network resources are allocated, but the system is not adaptive enough to accommodate diverse QoS needs of various applications and services
Solution Approach 1:
The patent utilizes QoS parameter negotiation and adjustment mechanisms that allow different applications and services to obtain appropriate QoS levels based on their specific requirements. The system modifies QoS parameters dynamically to accommodate diverse service needs while maintaining manageable complexity through standardized parameter sets
3Reliability
If network resources are over-allocated to ensure QoE requirements are met, then user satisfaction is maintained, but operational costs increase
Solution Approach 1:
The system implements partial QoS provisioning by allocating exactly the amount of network resources needed to meet user QoE requirements rather than over-allocating. Through precise QoS parameter control and dynamic adjustment, the network provides sufficient resources during high-demand periods while reducing allocation during low-demand periods, optimizing the balance between service quality and operational efficiency
Data Source
AI summary
A method for facilitating allocation of network resources to at least one slice in a communication network includes obtaining information identifying a mapping between network resource allocations and a respective quality of experience (QoE) associated with a user and obtaining information relating to a quality of service (QoS) tolerance for a particular QoE. The method includes providing to a controller for allocating at least a portion of the network resources to the at least one slice: i) the information identifying a mapping between network resource allocations and a QoE; ii) the information relating to the QoS tolerance; and iii) information identifying a current QoE and a target QoE associated with at least one user.