Daniel Jakobsson and Max Vahlström, MPSYS

​

Examiner: Knut Åkesson, Dept of Electrical Engineering

Supervisor: Kristofer Bengtsson, Dept of Electrical Engineering

The aim of this master thesis is to evaluate possible improvements of the perceived behavior of an Autonomous Mobile Robot (AMR) system using an RGB-D camera for Human Pose Estimation (HPE). This topic is approached by conducting an initial study of the current challenges and needs of humans working in shared local environments with AMR-systems. A study on how different HPE methods perform in an AMR system has also been conducted, which show that different methods could have very differing performance. Based on the results of the studies, a prototype implementation of some interaction concepts is presented, including the understanding of the human intentions as well as how an AMR should respond to these intentions.

The results of this thesis show that there are common issues for human users who are working along with current AMR systems. A proof of concept on how some of these issues could be solved is proposed, which potentially could improve the perceived behavior of the AMR significantly. Suggestions of HPE methods that have shown the most promising performance from tests is also given.