As robots increasingly integrate into our living environments, autonomous navigation in populated environments is getting an important challenge for robotics research. The key of the problem is to provide our mobile robots with the ability of moving in a reliable, safe, comfortable and natural way around humans.
Rapidly-exploring Random Trees (RRTs) have shown promising results in this context. In particular the algorithm RRT-x provides a fast replanning algorithm for unpredictably changing scenarios.
However, RRT-x, do not take into account the dynamics of human motion. Therefore, this thesis extends RRT-x with human motion predictions and a social cost function.
The developed system provides a safe, feasible, socially-aware and, eventually, time optimal path for the robot while accounting for unpredictable changes in the obstacle set of the environment as well as a real-time strategy to recover from pedestrian motion predictions failures.
The presented thesis have been developed in the “Social Robotics Lab – Albert-Ludwigs University of Freiburg im Brisgau” under the supervision of dr. prof. Kay O. Arras and of dr. Luigi Palmieri.