Scholarlink

<p>In certain unobserved environments with limited communication capabilities, the coordinated tracking of moving targets by multiple Unmanned Aerial Vehicles (UAVs) presents significant challenges. Due to the temporal decoupling of the search-allocation process, UAVs can only rely on delayed and incomplete historical observation data at decision-making moments. This results in significant probabilistic uncertainty regarding target states, rendering traditional task allocation methods based on static deterministic metrics (such as Euclidean distance) incapable of guaranteeing tracking feasibility and robustness. To address the above issues, this paper proposes an uncertainty-driven hierarchical distributed task allocation framework. First, based on a particle filter prediction model, the state distribution of the target dispersing over time is dimensionally reduced and represented as the re-capture probability, a sufficient statistic. Subsequently, a composite probability utility function integrating spatio-temporal reachability, remaining energy, and heading consistency is formulated, thereby realising the dynamic quantification of task value under physical constraints. During the pre-allocation phase, to address the consensus convergence challenge induced by dynamic utility, utility freezing and information credibility weights are incorporated into the Consensus-Based Bundle Algorithm (CBBA) framework, thereby realizing conflict-free provisional leader and target anchoring within a finite number of iterations. During the coalition formation phase, to address the stability challenges associated with many-to-one tracking, a local coalition formation mechanism based on marginal information gain is constructed. This paper theoretically demonstrates that the coalition formation process is equivalent to a precise latent game, employing coalition joint uncertainty as the potential function to ensure that distributed decision-making converges to a pure strategy Nash equilibrium. Simulation results indicate that this mechanism significantly enhances target capture rates and system robustness under conditions of information lag.</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

Uncertainty-driven collaborative target tracking: a hierarchical allocation mechanism based on probabilistic utility and latent games

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月13日

查看全文
  • Liyuan yang,
  • Yongping Hao,
  • Jiulong Xu,
  • Meixuan Li

摘要

In certain unobserved environments with limited communication capabilities, the coordinated tracking of moving targets by multiple Unmanned Aerial Vehicles (UAVs) presents significant challenges. Due to the temporal decoupling of the search-allocation process, UAVs can only rely on delayed and incomplete historical observation data at decision-making moments. This results in significant probabilistic uncertainty regarding target states, rendering traditional task allocation methods based on static deterministic metrics (such as Euclidean distance) incapable of guaranteeing tracking feasibility and robustness. To address the above issues, this paper proposes an uncertainty-driven hierarchical distributed task allocation framework. First, based on a particle filter prediction model, the state distribution of the target dispersing over time is dimensionally reduced and represented as the re-capture probability, a sufficient statistic. Subsequently, a composite probability utility function integrating spatio-temporal reachability, remaining energy, and heading consistency is formulated, thereby realising the dynamic quantification of task value under physical constraints. During the pre-allocation phase, to address the consensus convergence challenge induced by dynamic utility, utility freezing and information credibility weights are incorporated into the Consensus-Based Bundle Algorithm (CBBA) framework, thereby realizing conflict-free provisional leader and target anchoring within a finite number of iterations. During the coalition formation phase, to address the stability challenges associated with many-to-one tracking, a local coalition formation mechanism based on marginal information gain is constructed. This paper theoretically demonstrates that the coalition formation process is equivalent to a precise latent game, employing coalition joint uncertainty as the potential function to ensure that distributed decision-making converges to a pure strategy Nash equilibrium. Simulation results indicate that this mechanism significantly enhances target capture rates and system robustness under conditions of information lag.