Geometry-Aware Dual-LiDAR Pose Estimation of Power Lines for UAV Physical Interaction

Abstract

Inspection and maintenance of power lines are costly and hazardous operations that traditionally require human workers or helicopters. Uncrewed aerial vehicles (UAVs) offer safer and more flexible alternatives, but autonomous physical interaction with cables remains challenging due to the need for reliable relative pose estimation. This paper presents a dual-LiDAR perception system that estimates the relative position and orientation of power lines for UAV interaction tasks. The method uses two lightweight 2D LiDARs with parallel scan planes to detect cable intersections, associate observations across sensors in multi-cable scenarios, and estimate cable pose through an Extended Kalman Filter that explicitly models local cable geometry and sensor noise. The proposed system is invariant to illumination conditions and, unlike approaches that rely on sensing the electromagnetic field of the conductor, can operate on both energized and de-energized lines. The approach is implemented onboard embedded hardware and validated through indoor and outdoor experiments, including autonomous perching maneuvers and the deployment of a dual-arm manipulator on real power lines. The results demonstrate centimeter-level position accuracy and degree-level orientation accuracy, supporting the feasibility of UAV physical interaction with power-line infrastructure.