Training of construction robots using imitation learning and environmental rewards.
Saved in:
| Title: | Training of construction robots using imitation learning and environmental rewards. |
|---|---|
| Authors: | Duan, Kangkang1 (AUTHOR), Zou, Zhengbo1 (AUTHOR), Yang, T. Y.1 (AUTHOR) yang@civil.ubc.ca |
| Source: | Computer-Aided Civil & Infrastructure Engineering. 4/4/2025, Vol. 40 Issue 9, p1150-1165. 16p. |
| Subjects: | Observational learning, Imitative behavior, Gesture controlled interfaces (Computer systems), Robot design & construction, Machine learning, Reward (Psychology), Virtual reality, Learning |
| Abstract: | Construction robots are challenging the paradigm of labor‐intensive construction tasks. Imitation learning (IL) offers a promising approach, enabling robots to mimic expert actions. However, obtaining high‐quality expert demonstrations is a major bottleneck in this process as teleoperated robot motions may not align with optimal kinematic behavior. In this paper, two innovations have been proposed. First, traditional control using controllers has been replaced with vision‐based hand gesture control for intuitive demonstration collection. Second, a novel method that integrates both demonstrations and simple environmental rewards is proposed to strike a balance between imitation and exploration. To achieve this goal, a two‐step training process is proposed. In the first step, an intuitive demonstration collection platform using virtual reality is utilized. Second, a learning algorithm is used to train a policy for construction tasks. Experimental results demonstrate that combining IL with environmental rewards can significantly accelerate the training, even with limited demonstration data. [ABSTRACT FROM AUTHOR] |
| Copyright of Computer-Aided Civil & Infrastructure Engineering is the property of Wiley-Blackwell and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.) | |
| Database: | Engineering Source |
|
Full text is not displayed to guests.
Login for full access.
|
|
| Abstract: | Construction robots are challenging the paradigm of labor‐intensive construction tasks. Imitation learning (IL) offers a promising approach, enabling robots to mimic expert actions. However, obtaining high‐quality expert demonstrations is a major bottleneck in this process as teleoperated robot motions may not align with optimal kinematic behavior. In this paper, two innovations have been proposed. First, traditional control using controllers has been replaced with vision‐based hand gesture control for intuitive demonstration collection. Second, a novel method that integrates both demonstrations and simple environmental rewards is proposed to strike a balance between imitation and exploration. To achieve this goal, a two‐step training process is proposed. In the first step, an intuitive demonstration collection platform using virtual reality is utilized. Second, a learning algorithm is used to train a policy for construction tasks. Experimental results demonstrate that combining IL with environmental rewards can significantly accelerate the training, even with limited demonstration data. [ABSTRACT FROM AUTHOR] |
|---|---|
| ISSN: | 10939687 |
| DOI: | 10.1111/mice.13394 |
