Feedback-driven adaptive multi-robot timber construction

Abstract

Automation and robotics are anticipated to play a crucial role in addressing challenges confronting the construction industry, such as low productivity, workforce shortages, and physically demanding labor. However, a critical challenge in construction robotics has been the development of robust adaptive control to deal with uncertainties inherent in construction, such as material imperfections, multi-robot calibration, and fabrication inaccuracies. To address this challenge, we present a feedback-driven framework consisting of two complementary adaptive fabrication methods, pose-based and topology-based, incorporating perception, reasoning, and acting to handle uncertainties in multi-robot timber construction. We evaluate our framework through building-scale experiments, quantifying their deviations from their as-planned digital models. Our results indicate that our pose-based method significantly decreased deviations compared to a benchmark when applied to nail-laminated timber panels, and our topology-based method enabled robust multi-robot construction of a timber frame wall. Altogether, this research contributes to flexible, accurate, and robust construction employing multi-robot systems.