Conference Agenda

When optimizing transportation logistics through routing models, Euclidean distances are typically employed to test them and derive practical recommendations. However, transportation networks have varying topologies, especially in different urban environments. This leads to a discrepancy between Euclidean and actual distances, which poses a challenge to the applicability of these models in real-world scenarios. Spatial metrics could play a crucial role in bridging this gap between theoretical models and practical applications, as they help to understand the structures of transportation networks. They analyze the connections between different locations, provide information about the layout of roads, identify bottlenecks, and detect barriers. In this talk, we want to explore the relationship between certain spatial metrics and network efficiency in the context of routing problems. Specifically, pedestrian, bicycle, and road networks in different urban areas will be examined. First findings indicate that the approximation quality of Euclidean distances relies on the type of network and the size of the considered area. As expected, for localized trips, bicycle and pedestrian routes are more efficient than motorized traffic routing. However, there are exceptions, and the advantage of non-motorized vehicles varies depending on the characteristics of the network. It therefore seems useful to consider the urban structure before applying general results to practice.

As periodic timetables are used in public transport in many European countries, and in some cases also in freight transport, their construction, and optimization are inevitable tasks. We use the periodic event scheduling problem (PESP), first introduced by Serafini and Ukovich, as our underlying mathematical model. That PESP instances are generally notoriously hard to solve not only in theory but also in practice can be seen as none of the 16 instances of the benchmark library PESPlib could be solved to proven optimality so far. Therefore, focusing on the primal bound, we present a line-based primal heuristic in which we build an instance line by line. For this purpose, we define and analyze measures for the importance of a public transport line. Starting with the most important one, we optimize this subinstance and successively add further lines by decreasing order of importance. We propose a method to extend the solution of the previous iteration to the next to obtain a feasible starting solution by constructing a feasible periodic potential in a directed spanning tree and utilizing the instance structure. We then use the black-box solver ConcurrentPESP to optimize the solution for a fixed amount of time. Hereby, the starting solution is not fixed in the optimization step, so that previous decisions can be revised. In our computational experiments, the heuristic found better primal bounds for two out of the four tested PESPlib-instances. An improvement of 2.3 % was achieved for one instance.

In this work, we present a novel sim-learnheuristic approach to solving the Team Orienteering Problem (TOP), departing from the traditional focus on deterministic and stochastic variants of the task. By combining deterministic, stochastic, and dynamic features in a hybrid manner, our methodology presents a novel use of the sim-learnheuristic algorithm to address the complex problems of the TOP. Essentially, TOP is about distributing tasks and routing as efficiently as possible for a team while taking into consideration dynamic variables like weather, traffic, and resource variations. This task becomes much more complicated when stochastic travel times and dynamically changing factors are added. Our approach effectively navigates the dynamic travel times, forecasts future situations, and finds the best solutions by combining machine learning predictions, heuristic algorithms, and simulation-based optimization.

In Roll-on/Roll-off (RoRo) terminals, the maritime transshipment of vehicles is performed as a pre- and post-processing for maritime transportation with RoRo ships, collectively referred to as RoRo operations. RoRo operations are an important part of the vehicle distribution in automotive supply chains, and during loading and unloading, vehicles roll on and off a ramp to be stowed on the decks of the RoRo ship or stored in the parking areas of the RoRo terminal. In this context, the field of Operations Research offers the potential to address increasing operational costs and throughputs by supporting efficient and effective personnel scheduling. Since loadings mirror unloadings in reverse order, we focus on loadings for simplicity and consider personnel performing different operational tasks, such as personnel who drive vehicles from the parking areas to the decks, or personnel who drive shuttles to bring personnel from the decks back to the parking areas. Groups of vehicles stored in the parking areas and planned for loading are called batches. We consider the specific problem where there is a determined departure of the RoRo ship and there is a shortage of personnel in the RoRo terminal, so it is not possible to load all the batches as planned. We formalize this as multi-mode personnel scheduling and use a weighting for each batch to decide whether to load or not in order to maximize the sum of weightings of the loaded batches.