UAV Position Estimation using a LiDAR-based 3D Object Detection Method

TL;DR

The study addresses UAV position estimation in GPS-denied environments using LiDAR-only data from a UGV. It implements the PointPillars 3D object detector in MATLAB to identify the UAV and estimate its relative position, comparing against a traditional clustering approach. Results show the detector provides more frequent, centimeter-level X/Y updates and substantially improves detection coverage, though Z-axis accuracy is affected by unrecorded transforms in offline data; applying a transform correction aligns results closer to ground truth. The work demonstrates real-time viability and suggests future optimization, dataset expansion, and comparisons with alternative 3D detectors for improved robustness in cluttered scenes.

Abstract

This paper explores the use of applying a deep learning approach for 3D object detection to compute the relative position of an Unmanned Aerial Vehicle (UAV) from an Unmanned Ground Vehicle (UGV) equipped with a LiDAR sensor in a GPS-denied environment. This was achieved by evaluating the LiDAR sensor's data through a 3D detection algorithm (PointPillars). The PointPillars algorithm incorporates a column voxel point-cloud representation and a 2D Convolutional Neural Network (CNN) to generate distinctive point-cloud features representing the object to be identified, in this case, the UAV. The current localization method utilizes point-cloud segmentation, Euclidean clustering, and predefined heuristics to obtain the relative position of the UAV. Results from the two methods were then compared to a reference truth solution.

Figures (8)

• Figure 1: Position estimation using the clustering method
• Figure 2: Pointcloud representation of the tunnel with the UAV bounded by the green box
• Figure 3: LiDAR-based 3D object detection pipeline
• Figure 4: Overview of the PointPillars architecture
• Figure 5: Detection results of the network trained for 70 epochs (left); 150 epochs (right)