Constructive AI: autonomous robotic 3D printing enabled by AI-driven process control and 3D machine vision

Automation is a key promise of construction 3D printing (C3DP); however, achieving fully autonomous printing remains challenging due to the time-dependent material properties and complex process-material-environment interactions. Existing workflows rely on frequent manual interventions or post-deposition inspection. To address these limitations, this paper proposes an AI-based hybrid feedback-predictive methodology for autonomous and adaptive robotic layer deposition. Physics-guided features extracted from raw sensory data, including a specialized 3D vision module, material temperature, and rheological indicators derived from the extruder's electrical power consumption, were utilized to enable machine learning algorithms that estimate the optimum extrusion rate and interlayer delay under dynamic conditions. The proposed framework is deployed and validated through autonomous printing of two wall panels, achieving an average deformation of only 0.39% with a processing time of 0.6 s per output. The results confirm the effectiveness and strong potential of the proposed adaptive control methodology for achieving truly autonomous C3DP.