Combinatorial optimization offers a large range of classic benchmark problems. However, many real-world problems are often more complex in nature and bounded by further constraints. In this thesis, an autonomous robotic machine that batches chicken fillets into trays is studied. A combinatorial problem is solved in real-time by combining elements from different types of optimization problems. It looks at items currently available and seeks to pack them as efficiently possible while satisfying the necessary constraints.

Two models are proposed, GP and Hybrid, and they are compared to the existing algorithm currently used. The existing algorithm already has a good performance in practice, but it lacks transparency for the end users. The GP and Hybrid models are constructed as complex models, but as every step is well described and the number of parameters are minimal, they can facilitate a better understanding for the end users.

To facilitate a comparison with the existing algorithm, a simulation environment has been constructed that takes a variety of problem features. The two models prove competitive against the existing algorithm. The models are currently solved using CPLEX. Before deployment, it must be ensured that the processing time frame can be met, because currently they exceed the time limit about 5-10\% of the time. This can be done by either constructing a solution heuristic or by imposing certain rules.