Skip to main navigation Skip to search Skip to main content

RAFT: Relative-Pose-Aware Flexible Aerial Tracking of a Moving Target in Cluttered Environments

  • Beijing Institute of Technology

Research output: Contribution to journal › Article › peer-review

Abstract

In cluttered environments, maintaining a task-specified viewpoint during aerial tracking and preserving tracking continuity are crucial for applications such as inspection. However, occlusions and narrow passages can make the preferred relative pose temporarily infeasible, leading to viewpoint drift or loss of tracking continuity. This article proposes relative-pose-aware flexible aerial tracking (RAFT), a RAFT method that formulates the prescribed relative pose as a recoverable receding-horizon viewpoint reference. Specifically, the relative-pose-pattern (RPP) and tracking-continuity (TC) branches generate candidate viewpoint sequences over the same target-prediction horizon. The RPP sequence prioritizes the prescribed relative-pose reference, whereas the TC sequence provides rear-following viewpoints to preserve tracking continuity under temporary infeasibility. A predictive sequence-commitment mechanism triggers fallback before predicted safety or visibility violations and permits recovery only after sustained RPP feasibility is verified. A MINCO-based optimizer then converts the committed viewpoint sequence into smooth and dynamically feasible position and yaw trajectories. Comparative simulations and representative real-world experiments demonstrate that the proposed method maintains tracking continuity, preserves relative-pose consistency when feasible, and achieves smooth relaxation-and-recovery behavior in cluttered environments.

Original languageEnglish
JournalIEEE Transactions on Industrial Electronics
DOIs
Publication statusAccepted/In press - 2026

Keywords

  • Aerial tracking
  • motion planning
  • relative pose
  • visibility

Fingerprint

Dive into the research topics of 'RAFT: Relative-Pose-Aware Flexible Aerial Tracking of a Moving Target in Cluttered Environments'. Together they form a unique fingerprint.

Cite this