Material at manual working spaces is often organized in inflexible gravity roller conveyers. With every change of product variant the totes have to be rearranged manually. The goal of the dFlow project is the automation of this task. This system is based on the N x N puzzle but allows for more than one blank and multiple batch movements of blanks. The current algorithm is too slow for larger systems. The objective of this thesis is to reduce calculation time by employing heuristic search.
Different heuristic searches, which have been successful for the N x N puzzle, are examined in regards to their suitability for the dFlow system. The most promising search is implemented. The implementation is analysed and the heuristics are extended to match the differences of the dFlow system to the N x N puzzle. The optimized heuristics and search algorithms are then evaluated.
The developed searches are similar in regards to their average calculation time. Those times depend on the initial configuration of the system. The developed approaches return nearly optimal solutions. However, calculations with non-admissible heuristics are faster but return longer solutions.
The thesis shows that heuristic search is a promising approach to reduce calculation time. Also, the large influence of blanks in this system is discussed. To exhaust the potential of heuristic search the behaviour of blanks has to be considered in future heuristics.