Collision Avoidance Algorithm Using Gaussian Mixture Model and Vector Field Histogram 가우시안 혼합 모델과 벡터장 히스토그램+를 이용한 충돌 회피 알고리즘

Abstract

Obstacle collision avoidance is a key technology required for safe and reliable Unmanned Aerial Vehicle (UAV) operations. Since it is an essential factor of drone commercialization, several algorithms have been studied including Vector Field Histogram+ (VFH+) and Gaussian Mixture Model (GMM). In this study, an obstacle avoidance algorithm using the VFH+ and GMM algorithms is proposed against an environment with multiple obstacles. The VFH+ algorithm is adopted to design vector field histograms to determine the location and size of obstacles. And the GMM clustering is utilized to determine the density of obstacles. Based on these two methodologies, an avoidance direction is obtained and a path is created by applying the Dubins path planning algorithm. In order to compare and evaluate the proposed avoidance algorithm with existing collision avoidance techniques, a numerical simulation is performed based on a point mass UAV model with 3 degrees of freedom.