Data-based error compensation for georeferenced payload path tracking of automated tower cranes

In order to increase the productivity on construction sites, a current topic of research is the automation of the payload transport by tower cranes. Thereby, a key requirement is to ensure that the tower crane precisely tracks the planned paths and positions the payload at the specified target location. Most of the state-of-the-art tower crane controllers damp load sway while moving each driving system to its desired position. However, the path error also consists of bending displacements of the tower crane's mechanical structure, observer errors, or sensor offsets once the crane hook position is considered in a fixed georeferenced construction site system. These errors have not been addressed in literature on tower cranes so far. This paper introduces an approach to reduce the path error of automated tower cranes without permanently integrating additional sensors. A regression model is derived for predicting the path error and a least absolute shrinkage and selection operator (LASSO) is used to select the most important features. The predicted error is then used to compute a compensating hook path such that the measured hook path matches the desired hook path. The effectiveness of the approach is experimentally validated utilizing a real large-scale tower crane showing a reduction of the path error of more than 50% and a position accuracy of less than 16 cm.