Collision Avoidance for Multirotor Inspection

Abstract

During the last couple of years, the goal of using unmanned aerial vehicles (UAV) for inspection purposes has become within reach. This thesis considers the different algorithms for implementing a high level controller for an anti-collision system for a multirotor inspection drone operating inside a shipping tank. The desired platform for this thesis is a quadcopter UAV with six on-board radar sensor, providing environmental knowledge. A literature review of existing collision avoidance algorithms is performed to consider which ones that are suited for tank inspection. The linearized dynamic quadcopter mathematical model is derived and simulated with low level PID-control and high level collision avoidance in a self-developed MATLAB simulator. The methodology of the algorithms implemented is explained and divided into two groups; those considering the environment to be known (global), and the ones continuously sensing the environment from the on-board radar sensors (local). The Null-Space Based (NSB) behavioral control algorithm is tested as a global approach. A self-developed reactive logic algorithm and the Velocity Obstacle (VO) algorithm are tested as local approaches for collision avoidance. The results yielded from the different algorithms simulations show successful inspection with room for improvement on the implemented collision avoidance. The remarks of the simulation results are discussed and it is suggested to extend the self-developed reactive logic algorithm for future work.