LAUSR

Abstract Unlike manufacturing robots, whose kinematics are pre-programmed based on robust metrology, tight tolerances, and rigid workpieces, construction robots operate under conditions of imperfect metrology, loose tolerances, and large workpiece uncertainties. Despite having access to a designed Building Information Model (BIM), construction robots must sense and model their actual environment, and adapt their kinematic plan to compensate for deviations from the expected. This research investigates methods to enable the autonomous sensing and modeling of construction objects so construction robots can ultimately adapt to unexpected circumstances and perform quality work. To that end, two construction component model fitting techniques are presented, namely the Clustering and Iterative Closest Point (CICP) construction component model fitting technique and the Generalized Resolution Correlative Scan Matching (GRCSM) construction component model fitting technique. The GRCSM construction component model fitting technique employs the presented GRCSM search algorithm, which is a modified version of the existing Multi-Resolution Correlative Scan Matching (MRCSM) search algorithm. Three experiments are presented to evaluate the ability of the CICP and GRCSM construction component model fitting techniques to model construction features. It was found that the CICP and GRCSM construction component model fitting techniques are capable of estimating the pose and geometry of arbitrarily shaped objects and construction joints, but are susceptible to modeling error. Despite their limitations, the CICP and GRCSM construction component model fitting techniques appear to be promising tools for the geometric estimation of construction features, especially for situations involving full automation, detailed construction work, incomplete sensor data, and complex object geometry.