Mobile Robotics in Unknown Environments: Towards Full Autonomy

Mobile robotics has become increasingly popular in recent years as it provides an automated and cost-efficient solution to a variety of tasks. Traditionally, human operators would have full responsability on the robot actions with teleoperation. Recent advances in sensors and algorithms have paved the way for robots to be able to operate autonomously or with little human intervention. Autonomous operation in known and structured environments has been vastly studied over the last decades, but such scenarios are limited to specific and laboratory applications. Real world contexts are characterized by unknown and unstructured scenarios that the robot must sense and adapt while performing the prescribed task. Recently, great effort has been given to the development of strategies to face these challenges. However, the pursuit of full autonomy is hindered by the limited hardware capacity of mobile robots, that constrain the computational capacity available to realize the desired operation. In this thesis we present several strategies to cope with uncertainties and unknown environment for both ground and aerial robot mobility, with a particular focus on the efficiency and the compliance with real time constraints. First, we consider a task of robust robot coordination for object transportation. Then, a novel approach for reactive navigation in unknown environments is presented, with theroetical proofs and experimental validation. Additionally, we present a motion estimation algorithm for unknown environments with the purpose of aerial phyisical interaction. Overall, particular attention is given to the efficient implementation of the proposed methodologies, which is a key factor for achieving full autonomy.