[Submitted to IEEE RA-L] This repository is the official code for BESTAnP: Bi-Step Efficient and Statistically Optimal Estimator for Acoustic-n-Point Problem
Wenliang Sheng*, Hongxu Zhao*, Lingpeng Chen*, Guangyang Zeng†, Yunling Shao, Yuze Hong, Chao Yang†, Ziyang Hong, and Junfeng Wu
(* represents equal contribution.)
- [November, 2024] ReFeree is submitted to RA-L and paper is available on arxiv
- [November, 2024] Algorithm and simulation code is released
- A bi-step algorithm that solves the acoustic-n-point pose estimation problem (estimating sonar sensor's position and orientation using n matching pairs between 3D points in world frame and their 2D projections in sonar image) for 2D forward-looking sonar
- Decouples pose estimation into two steps: translation estimation via range-based localization, followed by rotation estimation
If you only want the source code for BESTAnP and baseline algorithms, you can go to folder MatlabCode.
In this work, we integrate our BESTAnP algorithm with feature point triangulation to develop a pure sonarbased odometry system. To the best of our knowledge, no study has put forward an odometry from sonar-only measurements.
Anp simulator is a ROS package developed under ros-noetic with which you can test your own algorithm. If you want to generate data set by our own, you can use our package.
- Lingpeng Chen ([email protected])
- Wenliang Sheng ([email protected])
- Guangyang Zeng ([email protected])