Publication: Development of an obstacle detection algorithm for unmanned aerial vehicle
Loading...
Date
2023-08
Authors
Ng, Harn Tung
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Recently, much attention has been given to Unmanned Aerial Vehicle (UAV) due to its wide range of applications that can benefit human beings. With the recent
advancement in chip fabrication technology, the compact and light mmWave Radar sensor has been used widely as perception sensor on UAV platforms. Typically, the mmWave Radar point cloud data obtained will be clustered using algorithms such as Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm to associate them to each of the obstacles present in the environment. However, the parameters for the DBSCAN algorithm are often fixed as a constant, without considering the varying flight environment of a UAV in each data frame. Furthermore, the DBSCAN algorithm does not consider the radial velocity information of each data point throughout the clustering process. All these issues can indeed cause the clustering result to be less precise. Due to that, this project presents a KNN-Probability method that can determine the 𝜀𝜀𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷 parameter of the DBSCAN algorithm dynamically for each of the point cloud data frame. Besides, a novel radial velocity based clustered refinement algorithm has also been proposed. Based on the analysis with the publicly available RadarScenes dataset, the proposed obstacle detection algorithm has achieved a better clustering precision with a high mean average precision (mAP) of 0.8942. Nevertheless, the proposed algorithms are sensitive to their respective parameters. Hence, the future research direction of this project will be
focusing on the development of artificial intelligence (AI) based methods to optimize
these parameters.