Towards Adapting Human Behaviour in Packing Logistics Through Imitation Learning

Despite numerous breakthroughs in the field of AI, there are still a large number of real-world application areas where humans cannot be replaced on an ad-hoc manner. Performance built through experience and knowledge is difficult to replicate, let alone surpass. Imitation learning offers an approach to make such strategies numerically accessible using demonstration data. In this paper we present an approach of transferring complex behaviour by experienced workers in packing logistics to a technical system. This method is investigated and evaluated on the basis of experiments motivated by real-life use cases. It is shown that imitation learning can be used to identify and describe characteristics of expert behaviour for this application. The intend is to use this as a basis for subsequent computation in especially volume minimization.