Joint Sleep, Power and RIS Control Across Network Timescales
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning algorithms for wireless network management struggle to handle control problems with different timescales, particularly in sleep and power control, which affects network performance and energy efficiency in heterogeneous networks.
Innovation Solution
A cooperative hierarchical deep reinforcement learning (Co-HDRL) algorithm and fractional programming (FP)-based method are employed for joint sleep, power, and reconfigurable intelligent surface (RIS) control, addressing timescale differences and optimizing network energy efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If conventional reinforcement learning algorithms are used for sleep control, then long-term dynamic optimization is achieved, but they cannot handle the timescale differences between sleep control and power control
Solution Approach 1:
The control algorithm is segmented into two distinct components: a meta-controller for sleep control operating on a long timescale, and sub-controllers for power control operating on a short timescale. This segmentation allows each controller to be optimized for its specific timescale, resolving the contradiction between handling multiple timescales and maintaining algorithmic simplicity.
Solution Approach 2:
The sub-controllers are nested within the meta-controller framework, where the meta-controller makes high-level sleep decisions and the sub-controllers handle detailed power control tasks. This nested structure enables the system to manage multiple timescales efficiently while maintaining a hierarchical organization that prevents excessive complexity.
2Loss of energy
If sleep control is implemented to enhance energy efficiency, then network energy efficiency is improved, but network performance may be degraded due to long-term decision impacts
Solution Approach 1:
The system implements feedback mechanisms where the meta-controller receives performance information from sub-controllers and adjusts sleep control decisions accordingly. This feedback loop ensures that sleep control decisions are made with knowledge of their impact on network performance, allowing the system to optimize energy efficiency while maintaining acceptable performance levels.
Solution Approach 2:
The sleep control decisions are made dynamic rather than static, allowing the meta-controller to adapt sleep schedules based on changing network conditions and performance requirements. This dynamic approach enables the system to balance energy efficiency improvements with network performance maintenance.
3Loss of energy
If reconfigurable intelligent surface (RIS) technology is deployed to improve energy efficiency, then energy consumption is reduced, but control of RIS phase shifts becomes complex in multi-BS and multi-RIS scenarios
Solution Approach 1:
The RIS control problem is segmented by associating each RIS with a specific base station, creating independent control units. This segmentation transforms the complex multi-RIS control problem into multiple simpler single-RIS control problems, reducing overall control complexity while maintaining energy efficiency benefits.
Solution Approach 2:
The fractional programming-based control method is designed to be universally applicable to any RIS-base station pair, regardless of the total number of RIS elements or base stations in the network. This universal approach simplifies control by providing a consistent methodology that works across different network configurations without requiring complex centralized coordination.
Data Source
AI summary
A method and network nodes for intelligent joint sleep, power and reconfigurable intelligent surface (RIS) control are disclosed. According to one aspect, a method in a network node configured to communicate with a plurality of small base stations (SBSs) includes jointly determining sleep control, transmission power control and reconfigurable intelligent surface, RIS, control for the plurality of SBSs based at least in part on a fractional programming (FP) algorithm, the FP algorithm configured to maximize a data rate for a plurality of wireless devices (WDs).