Kalibreringsmetoder för trafiksimuleringsmodell på mesonivå : Hur kan kalibreringsmomentet göras effektivare?

Sammanfattning: Advanced traffic models are used in traffic planning to predict the effects of different traffic actions. Common uses are to investigate the effect of new development or to assess the effects of various changes on the road network. To ensure that the traffic models' results are relevant to the site being modeled, they must be calibrated. A traffic model is no better than its calibration. However, this is also a very time-consuming task. The purpose of this report is to explain the various aspects required when calibrating a mesa model, as well as to provide concrete measures to reduce the time required for the calibration. This is done through a case study of a calibration in the Aimsun software at meso level. A literature study has been done which includes; traffic variations, traffic measurement methods, traffic models and calibration of meso models. By describing how traffic varies, and why, as well as how traffic is measured and the precision of the measurement methods, a background is given to the reality against which the model is calibrated. An overview of traffic models and how they are related to each other helps to understand what can cause errors in the models and an understanding of what it is that the model shows. The literature study ends with a review of calibration for meso models as well as different methods for quantifying the error in the model. A case study has been done in the Aimsun Next software. A large area has been calibrated against measured flows at the meso level with a total of 85 measuring points. First, traffic demand has been calibrated and then the network has been calibrated. In both steps, different methods are explained that can be used to reduce the time it takes to calibrate the model. A very good calibration result can be obtained if many different calibration methods are used. During the work it became clear that the key to shortening the time for the calibration is to work methodically and with a clear structure where different work steps must be performed in a certain order. The conclusion proposes a more detailed method. The visualization of the model was identified as a key factor in reducing the calibration time. The faster it is possible to see what is wrong in the model the faster it is possible to correct the error. It is therefore valuable to consider what needs to be visualized and when and how this can be done in the shortest possible time. Specifically for Aimsun Next, several different technical methods have been explained that reduce working hours for the most common tasks. Among other things, a method for identifying incorrect route choices has been presented in the conclusion. Using these methods, a very large model has been successfully calibrated. A regression analysis of the model and the control points gives an R2 value of 0.99, that is, the model corresponds to the measurement points with 99% accuracy.