Metaheuristic-based dispatching optimization integrated in multi-scale simulation model of railway operation

Abstract

The dispatching system serves as an integral component of railway operation control and aims to eliminate the negative impacts of unforeseen events occurred during the operation process. On account of the time-critical decision-making and the associated complexity of the process-integrated dispatching, an acceptable compromise must be found regularly between the processing time of a dispatching task, which is often determined by the computer technology, and the dispatching solution quality. This applies equally to simulation software of railway operation, in order to make the workload of operational investigations acceptable. Accordingly, it is particularly important in the design of dispatching tools to find a good balance between required computation time and sufficient quality of results. With this central goal, a dispatching optimization algorithm was developed in this dissertation, which is based on a widely used metaheuristic algorithm - tabu search - and the integration in a multi-scale simulation model. The approach is based directly on the findings from the DFG project “The influence of dispatching on the relationship between capacity and operation quality of railway systems” and expands these by a universal multi-scale model. The multi-scale simulation model is characterized by continuously scaling, in which railway operation processes are simulated on microscopic, mesoscopic and macro-scopic levels concurrently. For large investigation areas, the relevant areas are pre-sented on microscopic level, while the others are presented on more efficient mesoscopic and macroscopic levels. Furthermore, an assessment method for the multi-scale model was developed to determine the significant values of different in-frastructure elements in the investigation area. Depending on the significant values, the simulation model can migrate continuously between three abstraction levels, so that the computational complexity and the accuracy of simulation results are well-balanced. With the proposed multi-scale simulation model, the sequence of train movements can be determined by the simplest dispatching principle (First Come First Serve) or a predefined dispatching solution. “First Come First Serve” is employed to generate basic dispatching solutions, while predefined dispatching solutions are used to simulate and evaluate optimized solutions. The simulation model-supported tabu search-based algorithm for dispatching optimization is able to optimize the basic solution by a series of dispatching measures iteratively until a satisfactory solution is obtained. It could be proved by means of a reference example that the developed algorithm for dispatching optimization can provide a suboptimal/optimal solution in a limited time.