Deep Q-Learning Uplink Power Control for Cellular Interference
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In wireless cellular networks, managing uplink transmit power is challenging as high power causes interference to neighboring cells, while low power reduces throughput and may lead to coverage loss, necessitating dynamic adjustment of power control parameters to balance performance and interference.
Innovation Solution
Implementing a Deep Q-Learning Network to determine optimal values for power control parameters P0 and α for each cell, dynamically adjusting uplink transmission power based on traffic density, user distribution, and channel characteristics to maximize overall system performance without significant interference.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If transmit power is increased to improve uplink throughput and coverage, then system performance is improved, but interference to neighboring cells increases
Solution Approach 1:
The patent implements dynamic power control where the network entity continuously adjusts transmit power parameters based on real-time network state observations. The power control parameters are not fixed but adapt dynamically to changing traffic conditions, user distributions, and interference levels, allowing the system to optimize throughput while controlling interference through continuous adaptation rather than static settings
Solution Approach 2:
The patent employs feedback mechanisms where the network entity observes network state information including interference levels, traffic density, and user throughput metrics. These observations feed back into the power control parameter adjustments, creating a closed-loop system that learns from past actions and outcomes to optimize future power allocation decisions, balancing throughput improvement with interference control
2Object-generated harmful factors
If transmit power is decreased to reduce interference to neighboring cells, then interference is reduced, but uplink throughput and coverage are degraded
Solution Approach 1:
The patent applies local quality by allowing different power control parameters to be assigned to different users, cells, or network regions based on local conditions. Instead of uniform power reduction, the system selectively adjusts power parameters in specific locations where interference is problematic while maintaining higher power in regions where throughput is the priority, achieving spatially differentiated optimization
Solution Approach 2:
The patent changes power control parameters dynamically based on observed network state. Rather than fixed low power settings, the system adjusts parameters such as target received power, pathloss compensation factors, and power offsets according to real-time conditions, enabling the network to transition between interference-limited and throughput-limited operating regimes as needed
3Device complexity
If traditional power control methods are used with fixed parameters, then system complexity is low, but adaptability to changing network conditions is poor
Solution Approach 1:
The patent implements self-service through autonomous power control where the network entity independently observes network state, determines appropriate power parameters, and adjusts transmission without requiring complex manual configuration or centralized optimization. The system serves itself by learning from observations and automatically adapting parameters, reducing operational complexity while improving adaptability
Solution Approach 2:
The patent replaces traditional mechanical or rule-based power control systems with a learning-based approach. Instead of relying on pre-configured parameter tables or simple threshold-based adjustments, the system uses machine learning models that process network state observations and generate power control decisions, substituting complex mechanical adjustment mechanisms with intelligent algorithms that adapt automatically
Data Source
Figure 1
Figure 2
Figure 3A
AI summary
According to an aspect, there is provided a computing device for controlling terminal device uplink transmission power. Each terminal device is configured to determine uplink transmission power based on two power control parameters: a target received power for full pathloss compensation and a pathloss compensation coefficient. The computing device comprises means for performing the following. Information on data traffic in a plurality of cells is maintained in a database. The computing device initializes a deep Q-learning network in which a state is defined as cell-specific pairs of the power control parameters, an action is defined as a selection of valid values of power control parameters for a cell and a reward is calculated based on the information on data traffic. The computing device trains the deep Q-learning network to approximate a Q value function, determines optimal power control parameters based on thereon and causes transmitting them to access nodes.