Transport projects or studies that require UK government approval are expected to follow official transport analysis guidance (commonly referred to as TAG). In the newly released version 23 of Aimsun Next, it is now easier to create outputs recommended by the UK Modelling and Appraisal guidance from the Department for Transport (DfT) for monitoring changes in matrices. In this technical note, Tessa Hayman will explain the new features and how to use them.
In TAG Unit M3.1 Highway Assignment Modelling it is recommended that the changes from matrix adjustment (matrix estimation) should be monitored using:
- Scatter graphs comparing OD pair values before and after adjustment.
- Scatter graphs comparing origin destination totals (zonal trip ends) before and after adjustment.
- Comparisons of trip length distribution changes
- Comparisons of sector-to-sector level matrices before and after adjustment
Matrix adjustment shouldn’t make significant changes to the structure of matrices as these can cause significant changes in route choice or restrict the changes in route choices that may occur due to a scheme. It is therefore recommended that any adjustment is well supervised and the criteria for the significance of changes brought about by matrix estimation is given by TAG in the following table:
Matrix zonal cell values
Slope within 0.98 and 1.02
Intercept near zero
Matrix zonal trip ends
Slope within 0.99 and 1.01
Intercept near zero
R2 > 0.98
Trip length distributions
Means within 5%
Standard deviations within 5%
Sector to sector level matrices
Differences within 5%
Where the changes are greater than the significance criteria, then these should be explored by the modeller to understand if these will influence the purpose of the model. If they are deemed to be significant then the development of the prior matrix should be reassessed, the real data set checked. Any exceedances that are not deemed to be important should be explained and documented in the validation report for the model.
Matrix adjustment shouldn’t be allowed to make significant changes to the prior matrices, if calibration and validation targets are not able to be met without these changes then a lower level of validation should be reported.
To enable the modeller to monitor this more easily within Aimsun Next, several new features have been added. The first was added in v22, but should be highlighted, is the option to have a reference matrix that is different from your input matrix.
Within the static OD departure adjustment experiment and dynamic OD adjustment experiment windows, there is a dropdown that allows you to select the reference matrix for the adjustment. This is used to calculate the elasticities and constrain the matrix. This should be used when you are adjusting the matrix with multiple steps e.g., static OD adjustment then static OD departure adjustment and finally dynamic OD adjustment. It means that the adjustment will try to stay as close to the prior matrix used as input for step 1, rather than moving further and further away from the prior matrix at each step.