• obstacle avoidance
• stereo vision
• uav

Autonomous multirotor aerial vehicles have great potential to address and optimize solutions for many fields of applications. One of the essential requirements that characterize the architecture of an autonomous vehicle is the local path planner. This module is in charge of detecting and avoiding obstacles during navigation by acquiring sensor information from the surrounding 3D environment and computing collision-free path in real-time.
In this thesis, a three-dimensional collision avoidance algorithm has been implemented and tested in different unknown environments. After studying the advantages and disadvantages of algorithms in the literature, the 3DVFH+ algorithm has been chosen and implemented on a quadrotor equipped with a stereo-vision camera. The reactive control architecture allows the UAV to safely avoid obstacles during a mission plan with real-time sensor data. The algorithm has been developed in the ROS framework using software-in-the-loop paradigm. The effectiveness of the designed solution has been tested in Gazebo simulation environments where all the algorithm parameters tuning has been performed. Then experimental tests on a real quadrotor vehicle have also been done to prove both the feasibility of the proposed solution on a limited resources platform and the capability to work in real-life scenarios subjected to sensor noise. The proposed solution was also applied for the ICUAS 2022 UAV Competition, winning third place in the final phase.