Network Slicing via Actor-Critic Neural Network
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current edge computing systems face challenges in efficiently managing network resources to meet the diverse quality of service (QoS) requirements of heterogeneous wireless services, such as eMBB, URLLC, and mMTC, due to complex service level specifications and dynamic network conditions.
Innovation Solution
The implementation of a deep recurrent actor-critic machine learning-based network slicing solution that allocates resources using a partially observable Markov Decision Process (POMDP) and deep reinforcement learning (DRL), employing a neural network to make resource partitioning decisions and assess performance, with a critic network using long short-term memory (LSTM) architecture to learn from past experiences and improve resource allocation policies.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional network resource management is used, then system complexity is reduced, but QoS provisioning for heterogeneous wireless services deteriorates
Solution Approach 1:
The patent divides the network into multiple slices, each dedicated to specific wireless services (eMBB, URLLC, mMTC) with customized resource allocation and QoS parameters. This segmentation allows independent optimization for each service type without affecting others, resolving the contradiction between meeting diverse QoS requirements and maintaining manageable system complexity.
Solution Approach 2:
The patent implements dynamic resource allocation where network resources are flexibly assigned to different slices based on real-time service demands and QoS requirements. This dynamic approach enables the system to adapt to heterogeneous service needs without requiring complex static configuration for each service type.
2Productivity
If deep reinforcement learning with LSTM is implemented, then resource allocation performance is improved, but computational complexity increases
Solution Approach 1:
The patent pre-trains the deep reinforcement learning model with LSTM architecture offline to learn optimal resource allocation policies from historical network data. This preliminary action allows the model to capture complex temporal dependencies and service patterns beforehand, enabling efficient real-time decision-making without requiring heavy computational resources during actual network operation.
Solution Approach 2:
The patent introduces an intelligent controller as an intermediary layer between the network slice manager and resource allocation mechanisms. This controller implements the deep reinforcement learning model to translate high-level QoS requirements into specific resource allocation decisions, simplifying the overall system architecture while maintaining high allocation efficiency.
Data Source
AI summary
Systems, apparatus, articles of manufacture, and methods are disclosed to slice networks for wireless services. Example apparatus are to implement an actor-critic neural network to predict a quality of service metric for a network slice based on a long short-term memory representative of one or more prior slicing decisions, compare the quality of service metric with a target service level specification, and update the long short-term memory based on the comparison.