Scholarlink

<p>In pursuit evasion framework, managing tasks like chasing and tracking moving objects is complex due to the dynamic environmental conditions. Particularly this study focused on spacecraft task management scenario. Existing methods often fail due to the limited sensing abilities and inefficient real-time decision-making, leading to poor results in dynamic space applications. These issues arise because space environments are unpredictable and complex, and current systems struggle to adapt in real time. To address these limitations, we introduce the novel Dual-Bio Adapt Algorithm (DBAA) model, a novel bio-inspired approach based on natural predators’ behaviour, operated in two modes. The Eagle Vision Strategy (EVS) is applied to pursuers, and the Chameleon Confusing Strategy (CCS) is used for evaders. Following that, we implement the deep learning-based Multi-Agent Deep Reinforcement Learning (MADRL) approach to enable continuous learning for both pursuers and evaders to make optimal decisions in real time. Simulation results under spacecraft scenarios demonstrate that the proposed approach provides more efficient solutions for managing pursuit-evasion tasks.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A Novel Bio-inspired Optimised Algorithm with Deep Learning in the Pursuit-Evasion Framework for Task Assessment

  • 期刊论文

  • 发表日期：

    2025年5月30日

查看全文
  • Zheshu Jia

摘要

In pursuit evasion framework, managing tasks like chasing and tracking moving objects is complex due to the dynamic environmental conditions. Particularly this study focused on spacecraft task management scenario. Existing methods often fail due to the limited sensing abilities and inefficient real-time decision-making, leading to poor results in dynamic space applications. These issues arise because space environments are unpredictable and complex, and current systems struggle to adapt in real time. To address these limitations, we introduce the novel Dual-Bio Adapt Algorithm (DBAA) model, a novel bio-inspired approach based on natural predators’ behaviour, operated in two modes. The Eagle Vision Strategy (EVS) is applied to pursuers, and the Chameleon Confusing Strategy (CCS) is used for evaders. Following that, we implement the deep learning-based Multi-Agent Deep Reinforcement Learning (MADRL) approach to enable continuous learning for both pursuers and evaders to make optimal decisions in real time. Simulation results under spacecraft scenarios demonstrate that the proposed approach provides more efficient solutions for managing pursuit-evasion tasks.