MADDPG-Based Power Control Strategy for Unmanned Vehicles

In the domain of unmanned vehicle networks, efficient spectrum utilization is crucial due to the scarcity of spectrum resources. This paper addresses the challenge of dynamic spectrum access by proposing an advanced strategy to enhance spectral efficiency in such networks. We present a novel approac...

Full description

Saved in:
Bibliographic Details
Main Authors: Han, Zhenyuan, Sun, Lu, Wan, Liangtian, Luo, Ruoheng, Wang, Xianpeng
Format: Conference Proceeding
Language:English
Subjects:
Autonomous vehicles
Dynamic spectrum access
Echo State Networks
Heuristic algorithms
MADDPG
Power control
prioritized experience replay mechanisms
Solids
Spectral efficiency
Spectrum resource allocation
Stability analysis
System performance
Training
Vehicle dynamics
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:In the domain of unmanned vehicle networks, efficient spectrum utilization is crucial due to the scarcity of spectrum resources. This paper addresses the challenge of dynamic spectrum access by proposing an advanced strategy to enhance spectral efficiency in such networks. We present a novel approach that combines Multi-Agent Deep Deterministic Policy Gradient (MADDPG) with Echo State Networks (ESN) and prioritized experience replay mechanisms. Our approach focuses on optimizing the power control of unmanned vehicles to maximize overall spectral efficiency while maintaining communication quality for all vehicles. The proposed MADDPG_ESN algorithm integrates the ESN to effectively capture temporal dynamics and improve learning efficiency, and incorporates prioritized experience replay to optimize training stability and performance. This combination allows for rapid and accurate decision-making in the complex environment of unmanned vehicle communications. Our findings underscore the effectiveness of combining ESNs with prioritized experience replay in reinforcement learning for dynamic spectrum access, offering significant improvements in system performance and resource utilization.
ISSN:2770-2677
DOI:10.1109/SmartIoT62235.2024.00016