Adaptive Perception and Modeling for Robotized Construction Joint Filling

Construction robots must perceive and model their surroundings to compensate for uncertainties in their workpieces. This research investigates a technique to enable the autonomous sensing and modeling of construction objects and features so construction robots can adapt their work plans and perform work. To that end, the Generalized Resolution Correlative Scan Matching (GRCSM) construction component model fitting technique is presented, which registers BIM models to point cloud sensor data. The registration results enable the robot to update its workpiece models to reflect their actual condition. An experiment was conducted in which virtual sensor data was generated for a virtual construction joint, and joint profile models were registered to form a model of the joint. It was found that the GRCSM construction component model fitting technique can be used in combination with a lo w precision sensor to estimate the pose and geometry of a virtual construction joint with a mean norm positioning error of 1.7 mm. The GRCSM construction component model fitting technique appears promising for the geometric estimation of construction objects, especially for situations involving full automation, detailed construction work, incomplete sensor data, and complex object geometry.