An Integrated Heading-Aware Hierarchical Framework for UAV Exploration in Large-Scale Environments

Achieving comprehensive coverage and high exploration efficiency in large-scale post-disaster environments remains a critical challenge for autonomous UAVs. Existing methods exhibit three key limitations: decision oscillation induced by frontier fragmentation, efficiency degradation in large-scale scenes, and dynamic mismatch between planning and execution. To address these challenges, this paper proposes a three-layer tightly coupled hierarchical exploration framework integrating FCM-based adaptive frontier clustering, ATSP-based global viewpoint sequencing, and differential-flatness-based joint QP trajectory optimization. The central innovation lies in the heading-aware ATSP formulation, which explicitly encodes heading-change costs to enforce kinematic consistency between global planning and trajectory execution. Simulation experiments across four disaster scene categories and three spatial scales show that the proposed framework achieves an average coverage rate improvement of 15.5 percentage points in large-scale environments, reaching up to 26.9 percentage points, while reducing exploration time by 30% on average and up to 50%. Across 11 repeated trials, the proposed method maintains near-zero performance variance, in contrast to a spread of up to 46 percentage points observed in baseline methods, confirming strong reliability under diverse rescue scenarios.