Validation of robotic navigation strategies in unstructured environments: from autonomous to reactive

The main topic of this master thesis is the validation of a navigation algorithm designed to perform autonomously in unstructured environments. Computer simulations and experimental tests with a mobile robot have allowed reaching the established objective. The presented approach is effective, consistent, and able to attain safe navigation with static and dynamic configurations. This work contains a survey of the principal navigation strategies and components. Afterwards, a recap of the history of robotics is briefly illustrated, emphasizing the description of mobile robotics and locomotion. Subsequently, it presents the development of an algorithm for autonomous navigation through an unknown environment for mobile robots. The algorithm seeks to compute trajectories that lead to a target unknown position without falling into a recurrent loop. The code has been entirely written and tested in MATLAB, using randomly generated obstacles of different sizes. The developed algorithm is used as a benchmark to analyze different predictive strategies for the navigation of mobile robots in the presence of environments not known a priori and overpopulated with obstacles. Then, an innovative algorithm for navigation, called NAPVIG, is described and analyzed. The algorithm has been built using ROS and tested in Gazebo real-time simulator. In order to achieve high performances, optimal parameters have been found tuning and simulating the algorithm in different environmental configurations. Finally, an experimental campaign in the SPARCS laboratory of the University of Padua enabled the validation of the chosen parameters.