Combining motion planning with social reward sources for collaborative human-robot navigation task design

Abstract

Across the human history, teamwork is one of the main pillars sustaining civilizations and technology development. In consequence, as the world embraces omatization, human-robot collaboration arises naturally as a cornerstone. This applies to a huge spectrum of tasks, most of them involving navigation. As a result, tackling pure collaborative navigation tasks can be a good first foothold for roboticists in this enterprise. In this thesis, we define a useful framework for knowledge representation in human-robot collaborative navigation tasks and propose a first solution to the human-robot collaborative search task. After validating the model, two derived projects tackling its main weakness are introduced: the compilation of a human search dataset and the implementation of a multi-agent planner for human-robot navigation