Autonomous navigation in unstructured environments using an arm-mounted camera for target localization

Autonomous mobile robots became over the recent years a popular topic of research mainly for their capacity to perform tasks in complete autonomy without the constant intervention of an human operator. In this context, autonomous navigation represents one of the main studied branch of autonomous robotics. Autonomous navigation in both structured and unstructured environments have been widely researched over years, with the development of several techniques that tries to solve this problem. In this context, there are several components that are required to get the proper solution to the navigation problem, and one of these is represented by the knowledge of the final position that an autonomous robot has to reach inside an environment. In this thesis, the goal is to enhance the autonomous capabilities of a robot by making it able to detect and follow constantly a target placed inside an unstructured environment. This result is obtained using a camera installed as end-effector of a robotic arm, which in turn is installed on top of a mobile robot. All the methodologies as well as the tools that have been used in the development of this project are presented in this thesis. The evaluation of the performances of the algorithm are performed both in a static context, where the robot is fixed and the target is free to move, and in a dynamic context where the robot moves and the target is fixed. The motion of the robot is obtained using an innovative algorithm for navigation in unstructured environments, NAPVIG. The proposed approach has been implemented using ROS and been tested both in a simulated environment using Gazebo as well as in a real world scenario. The results obtained from both type of experiments will be presented and discussed.