Treffer: 基于 NRO-QMIX 的战场环境 多无人机协同目标搜索.

To address the challenges of target search and information gathering in unknown battlefield environments, this study proposes an NRO-QMIX multi-UAV cooperative search method based on a multiagent reinforcement learning framework. First, task environment, target, and UAV models are constructed to accurately represent real-world mission scenarios. A regional information credibility map integrating target search and information gathering tasks is introduced, which transforms the multi-UAV cooperative search problem into a credibility map optimization problem. To overcome the low exploration rate and slow convergence in QMIX, the NROWAN noise network is incorporated, forming the improved NRO-QMIX algorithm, which enhances agent exploration and accelerates convergence. Additionally, an information-rich UAV state space is constructed based on the partially observable Markov decision process (POMDP) characteristics of multi-UAV cooperative search, and a task-specific reward function is developed to further boost convergence speed of NRO-QMIX algorithm. Simulations using Python show that the NRO-QMIX algorithm significantly outperforms traditional QMIX and Random algorithms in terms of [ABSTRACT FROM AUTHOR]

针对战场未知环境下的目标搜索与信息收集任务, 提出一种基于多智能体强化学习框架 的 NRO-QMIX 多无人机协同搜索方法。 建立任务环境模型、目标模型和无人机模型, 以准确描述 实际任务场景; 提出一种集成目标搜索与信息收集任务的区域信息可信图, 将多无人机协同搜索求 解问题转变为优化可信图问题; 针对 QMIX 算法引导的无人机搜索率低、求解速率慢的问题, 引入 NROWAN 噪声网络, 提出改进的 NRO-QMIX 算法, 以更好地引导智能体搜索并加速求解过程; 基 于多无人机协同搜索的部分可观测马尔可夫决策过程特性, 提出信息丰富的无人机状态空间构建 方法, 并基于任务需求与环境特征构建奖励函数, 以增强 NRO-QMIX 算法收敛速率。 利用 Python 软件进行仿真验证的结果表明, 与传统 QMIX 算法、Random 算法相比, NRO-QMIX 算法收敛速度 具有明显优势, 并在目标搜索与信息收集任务中展现出更优的性能。 [ABSTRACT FROM AUTHOR]

Copyright of Journal of Shandong University of Technology (Natural Science Edition) is the property of Shandong University of Technology and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)