Spatial Referencing for Human-Robot Collaboration in Construction

1. We propose a new interface which can provide an effective and intuitive means of operating construction robots, enabling construction workers to communicate spatial goals without the need for extensive training in robotics. Current construction robots are limited in their ability to understand spatial instructions and goals on the worksite. Moreover, user interfaces, such as handheld controllers, require extensive training and operation time for users, which can be impractical for construction workers who are typically not experts in robotics.

2. We propose a method to accurately infer the distant, human-indicated spatial goals by estimating the laser trajectories. Despite the advantages of deictic gestures, current gesture interfaces may not be suitable for construction tasks that require a high level of precision. Within large-scale environments, these interfaces cover only a restricted proximity of the entire workspace and have limited accuracy in estimating the spatial locations of distant targets. Interfaces from other domains are mostly tested on close-proximity tabletops, whereas this study focuses on scenarios with more distant targets. This underscores the relevance and applicability of the proposed method in real-world scenarios where construction robots operate within expansive 3D workspaces.