Mixed human-robot teams are increasingly considered for accomplishing complex mission due to their complementary capabilities. A major barrier for deploying such heterogeneous teams in real-world settings, is the current lack of natural skills in robotic team members, such as the understanding and interpretation of natural language instructions that include referential descriptions of entities in the world. In this paper we report the results of an empirical study in which humans tend to use referring expressions. We show how the received results and ideas can be used as guidelines to improve dialogue systems. By integrating and extending our system with these results, we will show how complex natural language instructions can be easily translated by robotic systems.
@inproceedings{yazdanietal17cic,
title={Guidelines for Improving Task-based Natural Language Understanding in Human-Robot Rescue Teams},
author={Yazdani, Fereshta and Scheutz, Matthias and Beetz, Michael},
year={2017},
booktitle={Proceedings of the 2017 8th IEEE International Conference on Cognitive Infocommunications},
url={https://hrilab.tufts.edu/publications/yazdanietal17cic.pdf}
doi={10.1109/CogInfoCom.2017.8268243}
}