Underwater Relative Pose Estimation and Tracking Between AUVs Based on a Hull-Integrated Laser-Beam Marker

Close-range tracking between autonomous underwater vehicles (AUVs) requires continuous six-degree-of-freedom (6-DoF) relative pose feedback, yet target boards and protruding light arrays are difficult to integrate into compact streamlined hulls. This paper presents a monocular relative pose sensing and tracking system based on a hull-integrated laser-beam marker: four blue beams emitted through hull-flush outlets form four calibrated coplanar intersections. A Hough-strip method recovers the principal beams and their intersections from fragmented, merged, and spurious image responses. To suppress the pose jumps that image-point errors induce through the coplanar ambiguity, the proposed solver refines multiple planar candidates and, in each frame, outputs the candidate consistent with both the current observation and the preceding motion. The system runs onboard in real time. In fixed-pose pool tests over 2-6 m, the lateral and attitude errors were at the centimeter and degree levels, respectively, and the normalized optical-axis root-mean-square error (RMSE) was 2.4-2.8%. On dynamic lake data, the proposed solver did not exhibit the large attitude jumps observed for representative perspective-n-point (PnP) solvers and reduced the root mean square of the third-order attitude differences to about 1/11 of the single-initialization value. In a 214 s nighttime lake trial, 99.44% of frames yielded valid poses, supporting three-channel closed-loop tracking.