Start-colloquium Maikel Brinkhoff

The engineering and construction (AEC) industry aim to increase its efficiency and output using concepts and technologies of Industry 4.0. The applicability of one such technology, on-site robotics, is researched in this work. A construction site is dynamic and therefore introduces many uncertainties. One example of uncertainty is (the use of) construction materials: there can be geometric deviations or sudden condition changes imposed by the weather. This work aims to design and construct a structure (e.g., arch or column) with irregular and varying geometries to mimic said geometric deviations. Such a problem is NP-Hard, meaning it cannot be solved in polynomial time by regular algorithms. To that end, Deep-Q Learning is used as a sequential decision-making algorithm to approximate the best solution. Verification of the solution is done by robotically constructing the structure.