Abstract:

In the paper, we introduce a novel framework for combining two modelling techniques: agent-based models (ABM) and Multi-resolution modelling (MRM). Both of these methods are well known and wide practical usage. However, their use in a single environment, and in addition reinforcement learning, provides additional opportunities for a variety of (in terms of conditions) test scenarios, shortening decision verification time, obtaining synthetic simulation data for further research and optimisation for uncharacteristic conditions. By using the approach it is possible to analyse many business strategies (e.g. what would happen if shipments were delivered twice a day).