Abstract (eng)
This thesis studies a temperature-dependent and multi-objective Capacitated Vehicle Routing Pickup and Delivery Problem, titled the TD-CVRPDP, that stems from a logistic steel plant application. In this routing problem, crude items need to be transported within a steel plant by heavy-duty vehicles. These vehicles are subject to the items' high temperatures and heavy weights. The routing problem at hand is the first of its kind to consider item and vehicle temperatures and their interdependence. Hence, this thesis conducts fundamental research on both the determination of the vehicle temperature during route planning, i.e., by means of the so-called Vehicle Temperature Prediction Method based on Newton's Law of Cooling (NLC) that was developed through inductive reasoning in the course of multiple on-site experiments at the partnered steel plant, and the design of a solution framework capable of handling the temperature aspect in a holistic manner, i.e., by means of a customised Adaptive Large Neighborhood Search (ALNS) metaheuristic that incorporates the temperature aspect as part of the objective function and neighbourhood evaluation. The consideration of the fleet's temperature is accompanied by difficulties regarding unexpected side effects in the solution structure, challenging the solution's practical applicability. Hence, the so-called Directional-Penalty is introduced, a penalty function applicable to a wide range of routing problems that, briefly put, aims to minimise the travel time of the items' individual journeys. Additionally, a new heuristic, titled the Dynamic-5-Request-Look-Ahead-Tree-Search (D-5R-LA-TS), is presented, designed to solve the TD-CVRPDP as a dynamic problem. Finally, all applied solution methods, conventional and new, are explained in great detail and thoroughly related to existing methods in the literature. Computational experiments demonstrate the effectiveness of all proposed solution methods and reveal valuable managerial insights by analysing the solution structure of the TD-CVRPDP as opposed to the generalised CVRPDP.