Deep Reinforcement Learning-based Construction Robots Collaboration for Sequential Tasks

摘要

Deep Reinforcement Learning-based Construction Robots Collaboration for Sequential Tasks Lei Huang and Zhengbo Zou Pages 48-51 (ICRA 2022 Future of Construction Workshop Papers, ISBN -, ISSN 2413-5844) Abstract: The integration of robots into the construction industryshows promise in addressing challenges such as stagnantproductivity and low efficiency. Recently, an increasing amountof research develops construction robots based on reinforcementlearning (RL). However, most existing RL-based constructionrobots are trained to conduct specific tasks individually withoutcooperation. This paper proposes an approach that utilizes twoRL-based construction robots (an unmanned ground vehicleand a robot arm) to collaboratively finish the task of windowpanel transport and installation in sequence without humanintervention. Our experiment results show that the two constructionrobots can successfully collaborate to finish all tasksin an end-to-end manner after they are trained separately witha success rate of 79.6%. Keywords: No keywords DOI: https://doi.org/10.22260/ICRA2022/0015 Download fulltext Download BibTex Download Endnote (RIS) TeX Import to Mendeley