Dates: 2023
Principal Investigator: Dr. Hussein bin Samma
Description: The UAV industry has seen a rise in the number of businesses employing UAVs for visual inspections. This is due to many reasons such as its efficiency, low cost, and the effectiveness to inspect at heights as well as inaccessible areas. This research aims to develop a visual navigation system based on deep reinforcement learning which will be applied for autonomous inspection applications. The implemented system will consist of several components, including the navigation agent based on deep reinforcement learning, deep learning vision models for scene encoding, obstacle avoidance model, object detection model to localize the defected areas during the inspection task, path planner to generate various routes during the inspection task, and a reward function to evaluate agent actions. This project will be implemented in a simulation environment using different tools such as Microsoft AirSim, ROS, Gazebo, Webots, and Mujoco. As a case study for inspection, the created system will be applied to the examination of transmission lines for energy.