Online Control for Same Day Delivery with Mixed Autonomous Fleets

This dissertation aims at controlling the logistics of same day delivery on the level of short termed decisions. Stochastically arriving customer requests are assigned to a mixed fleet such that a) delivery time windows are kept and b) future customer requests can still be handled adequately well due to anticipative fleet management. This problem is to be modeled as a dynamic stochastic model and will be solved by methods of approximate dynamic programming. This approach combines optimization, simulation and machine learning. With respect to the application mixed fleets are considered comprising autonomously acting vehicles as well as ones requiring manual interaction. The delivery decisions to be taken shall consider the attributes of the vehicles of a mixed fleets in an adequate way.