An evidence‐based approach to object pose estimation from LiDAR measurements in challenging environments

Abstract

This paper addresses the problem of estimating object pose from high-density LiDAR measurements in unpredictable field robotic environments. Point-cloud measurements collected in such environments do not lend themselves to providing an initial estimate or systematic segmentation of the point-cloud. A novel approach is presented that evaluates measurements individually for the evidence they provide to a collection of pose hypotheses. A maximum evidence strategy is constructed that is based in the idea that the most likely pose must be that which is most consistent with the observed LiDAR range measurements. This evidence-based approach is shown to handle the diversity of range measurements without an initial estimate or segmentation. The method is robust to dust. The approach is demonstrated by two pose estimation problems associated with the automation of a large mining excavator. K E Y W O R D S lidar, mining automation, perception, pose estimation, sensors 1992), and many have pursued using similar metrics. Blais, Beraldin, El-Hakim, and Cournoyer (2000) minimize the quadratic error J Field Robotics. 2018;35:921-936.

Real-Time 6-DOF Pose Estimation of Known Geometries in Point Cloud Data

“…The Maximum Sum of Evidence method previously demonstrated by the authors in [ 2 ] is not displayed due to its high computational expense, unable to satisfy the timeliness criterion. However, the results in Section 6 are compared against the original dataset used to demonstrate MSoE in [ 11 ]. All testing was conducted on an Intel i7 CPU @ 3.80 GHz with 62.5 GiB memory and an NVIDIA GeForce GTX 3090 GPU running Ubuntu 20.04.4 LTS.…”

“…Many existing algorithms require significant preprocessing, configuration tuning, or complex geometric calculations, and are unable to provide a real-time solution. For example, our previous work in [ 11 ] presented a robust and accurate pose estimation method with a high computational expense. We reported 99.5% of the total execution time being utilised by the raycasting operation in the objective function.…”

“…Borthwick demonstrated a method for haul truck estimation and load profiling using stereo vision and the ICP registration algorithm [ 23 ], whereas Cui et al [ 24 ] presented a registration network for excavator bucket pose estimation. Phillips et al demonstrated a method for both haul truck and dipper pose estimation using a LiDAR sensor and the Maximum Sum of Evidence method [ 11 ]. In the various applications mentioned, algorithms tend to focus on either the solution’s robustness or the solving efficiency.…”

LiDAR-Stabilised GNSS-IMU Platform Pose Tracking

“…Refs. [ 35 , 36 ] show how pose estimates of known objects can be obtained from only LiDAR scans by finding the pose that is most likely among the set of all possible poses.…”

Maximum Sum of Evidence—An Evidence-Based Solution to Object Pose Estimation in Point Cloud Data

“…The method of this paper, which we call Maximum Sum of Evidence, overcomes this limitation by seeking to determine the pose that is most evidenced in the point cloud. The paper builds from [ 9 , 10 ] and demonstrates the efficacy of an evidenced-based approach to addressing ‘ where is it? ’, ‘ what is it?…”