Railway Container Scheduling With Simulation Based Optimization Approach

Container transportation has become an indispensable part of logistics system since it features the merits of large freight capacity, convenient transportation, and high safety, etc. The scheduling of container transportation has attracted extensive attention of scholars due to its importance in the national economy.At present, worldwide scholars habitually pay more attention to the container transportation in the port, but less to the railway container transportation. Compared with the port system, railway container transportation has the advantages of a fast turnover, short-cycle loading and unloading, high quality of service requirements. Therefore, the higher quality scheduling of container regulation is required. Meanwhile, container scheduling in both railway and port are complex stochastic systems, which are difficult to optimize the entire process based on mathematical programming methods. Consequently, in this paper, Simulation Based Optimization (SBO) is presented to optimize the container handling process in the railway container terminal.SBO is a promising choice to solve the optimization problems which are hard to be described by mathematical models. Due to the high computational cost including both CUP time and memory space, it is difficult to apply SBO into practical problems. The reasons are presented as followings. First, the output of simulation model is used to replace that of the objective function for SBO method, and the running time of the simulation model is much larger than the calculation time of the mathematical function. Second, due to the randomness of the problems, the simulation has to be executed many times for obtaining a statistic evaluation value, resulting in the calculation time doubled.A simulation model regarding railway container scheduling is constructed, by which, gantry scheduling, container truck scheduling and some other scheduling strategies are researched. The route planning strategy of the container truck is also investigated. Considering the phenomenon that the paths in the railway container terminal are narrow and limited, the idea that the path resources act as the study objective is put forward, and Euler Equation is applied to implement the idea. It makes the container truck has the ability of predicting the congestion trend on the path from two aspects of time and space. Then the Euler factor and the distance factor are both thought of to make the container truck choose less congested and shorter path as its route.In order to save calculation, several approaches which can avoid simulation or decrease simulation cost are presented and applied:(1) Virtual Evaluation:We refer a solution as a dominant non-optimal solution once it can be determined as a bad one by its own information. The evaluation of solution, which does not rely on running the simulation model but according to the solution itself, but is obtained from a simple formula, is called as Virtual Evaluation. A large number of the simulation iterations, which are ascribed to lower-quality solution calculation effort, can be avoided by the virtual evaluation method.(2) Case Searching:The solutions obtained from the simulation process are stored in the case base in sequence of their evaluation value from good to bad. With a view of subsequent computation indexing, the stored previous computation results can be reused. The cases stored in the case base are updated with the calculation iterations, including the solution, the solution evaluation value, the order of the solution, the number of simulation times and so on. Case searching method can avoid duplication of simulation. The case base is the realization platform of several other methods.(3) Adjusting Simulation Run-Numbers:The simulation run-numbers of the solutions can be allocated in accordance with its current evaluation. The better solution would be allocated to the larger number of simulation so as to get evaluation as precisely as possible. The worse solutions are unlikely to be the optimal ones, so less number of simulation are assigned. Less number of simulations can save a large amount of computing costs. By using the method of adjusting simulation runs, the appropriate allocation of computation resources can be implemented.(4) Adjusting Step Length:The time step length is allocated to the solutions depending on their evaluations. Since the smaller the step length, the higher precise of the simulation results are achieved. However, the consumption for computation cost is also larger. Therefore, the smaller step length should be allocated to the solution with the better evaluation, and the longer step is assigned to the inferior ones. By using this method, the appropriate allocation of computation resources can also be implemented, and the computation resource are spent on the superior solutions.(5) Simulation Break:During one simulation process, although the simulation has not yet been completed, the final output of a complete simulation can be approximately estimated by the current output, thus, the simulation can be broken. The method of simulation break set up a number of break points in the simulation program. At each break point, once the difference between the output of current solution and that of the best solution in the case base breaks a threshold, the simulation program is terminated at that point immediately, and the evaluation of the simulation is estimated according to the output of the simulation which has been partly operated. For an inferior solution, the earlier the break point is, the more calculating cost is saved.The methods investigated in this dissertation are implemented to the railway container terminal to optimize the container schedule strategy. Based on the optimization results, the container schedule strategies are also analyzed and concluded.