Thomas Salmikeit

Integrated Planning of Delivery Traffic under Uncertainty

PhD studentThomas Salmikeit
Project startAugust 1, 2005
FundingVolkswagen Logistics


The logistical performance of industrial and commercial enterprises is one of today's decisive competitive factors. Current and future challenges of logistics consist in cross-departmental and cross-company logistics optimizations. The development of system partnerships, the integration of logistic service providers as partners in supply networks as well as the process integration through information and communication systems are important success factors. For companies in the automotive industry, the process of delivering material between the involved process participants forms an interface that can be further optimized in the context of the integration of logistic service providers and process integrations. Thus, this work focuses on the optimization of integrated planning of delivery traffic under uncertainty.

A central component of integrated planning of delivery traffic under uncertainty is the creation of robust time window plans as a connecting element between the route planning and the operative truck control on the recipient's premises. The mathematical model needed to compute robust time window schedules was developed based on models of resource-constrained project scheduling. The chosen robustness measure corresponds to a distribution-free approach, with no assumption of knowledge of the probability distributions for the uncertain execution durations of the activities. After a review of alternative approaches to robust project scheduling, the order theoretic approach was identified as a suitable approach for the problem at hand. Resource conflicts that arise can be determined and subsequently resolved in a performant manner by applying network flow algorithms through a suitable graph-theoretic interpretation.

The algorithms developed for the integrated planning of delivery traffic under uncertainty were analyzed experimentally with respect to their performance. Here, the influence of different parameters on CPU runtimes and robustness values on the one hand and the quality of time window schedules on the other hand were investigated. A central result of this work is that the robustness measure is a very good indicator for the actual robustness, despite its conceptual simplicity. Another key result is that the use of robust time window schedules in integrated scheduling of delivery traffic under uncertainty leads to significantly fewer delays in dispatching trucks at the consignee's premises compared to the conventional time window schedules used today.