Stochastic assignment: Basis tab
|
Element |
Description |
|
Use current assignment result as initial solution |
Note This option is recommended in the case of modified demand (e.g. due to 'Read matrix additionally' or matrix editing in the matrix editor). First, the former assignment result is scaled according to the new demand and then the assignment continues to calculate the new result. Usually, this requires less iterations than a complete recalculation. In this way, a lot of computation time can be saved in models with feedback from the assignment calculation to the demand calculation.
|
|
Global termination condition |
Maximum number of iterations The procedure terminates after the specified number of external iterations. |
|
Termination condition for network balancing |
Maximum number of iterations Via this button, you can set the number of balancing iterations. The number may vary due to the fact, which of the external iterations is currently being calculated. This is why you can assign the steps of external iterations to classes (Setting the number of balancing iterations). Notes In the first external iterations, in order to save computation time it might be useful to calculate a lower number of internal balancing iterations prior to searching for new paths. The button label shows the current settings in brief. Parameters in formulas The procedure terminates as soon as the deviation of impedances and route volumes calculated in two subsequent iteration steps goes below a certain threshold. You may adjust the formula for the threshold. |
If this option is checked, the assignment will use existing paths and volumes of the previous assignment as initial solution.
If this option has not been checked, the calculation starts from the unloaded network.Stochastic assignment: Impedance tab
For each demand segment chosen for the procedure, there is an Impedance tab. By clicking on the general impedance symbol, if necessary, you can set an additional special impedance for each demand segment via the complete path. You can, for example, model bicycle or heavy goods transport better. Also, the general impedance for each transport system which is displayed in the first line is used.
|
Element |
Description |
|
Coefficient |
Factor by which the impedance is multiplied. |
|
Property |
You can insert a description here. |
|
Formula |
Specify here the formula for the user-defined impedance for each path for the relevant demand segment. For bicycle traffic, for example, you can enter here the maximum slope [MAX:LINKS\SLOPE] on the entire path. |
Stochastic assignment: Smoothing tab
The parameters correspond to the parameters of the Equilibrium_Lohse assignment (Parameters of the Equilibrium_Lohse assignment).
|
Notes: In contrast to the Equilibrium_Lohse assignment, the value range of the exponent is 0 – 2 in the Route volumes section. For smoothing the route volumes, we recommend an exponent between 1 and 1.5. |
Stochastic assignment: Search tab
In this tab, you define how to determine PrT routes.
|
Element |
Description |
|
Randomized search |
Execute If this option has been checked, the best-path search is performed several times, each time using different conditions: link impedances are randomly increased or reduced In this way, also less attractive routes are found. This option models the differences in the road users´ ratings. Number of extra search iterations Via this button, you can set the number of extra search iterations for finding more routes. The number may vary due to the fact, which of the external iterations is currently being calculated. This is why you can assign the steps of external iterations to classes (Setting the number of search iterations). Notes For saving computation time it might make sense to calculate a different number of extra search iterations in the first internal balancing iterations. The button label shows the current settings in brief. By default, no classes have been defined, thus, the number of search iterations is the same for all external iterations. Sigma Standard deviation of the normal distribution (with mean R‘) from which the randomly varied impedance is drawn. Note The power of the impedance R‘ goes into the calculation. An exponent of > 0.5 leads to increased scattering and thus to more alternative routes. |
|
Random variation of the coefficients of the user-defined impedance components |
Execute
If this option is selected, the search is also performed with normal distributions of faded impedances on network objects. This has the effect that several different paths remain for varying the coefficients. For this, in addition to the true coefficient sentence β_0,β_early,β_late,β_1,…,β_n additional coefficient sentences are used, for which the β_1,…,β_n come about by fading the original values. The coefficient sentences take effect only for the detour test and preselection. Note The option is available if you have defined user-defined impedances in an Impedance tab. |
|
Detour test |
Execute Select this option to discard paths that match existing paths, except for one segment that is considerably longer than in the existing paths (Fundamentals: The procedure of stochastic assignment). A path P is deleted if the following applies to an existing path P’ for a mesh Via this formula you define how much longer the travel time of a route segment may be at the most before the longer path is discarded. Here, the travel time t0 in the unloaded network is used. and the entire impedance of the path P is larger than the impedance of P' with respect to all coefficient sets If this option is selected, an underlying path P is not discarded if it has, for at least one coefficient sentence, a better path evaluation at the demand segment level. |
Setting the number of search iterations
1. Make sure that in the Parameters: Stochastic assignment window, the Search tab is open (Stochastic assignment: Search tab).
2. Click the button next to Number of extra search iterations.
The Number of extra search iterations window opens.
|
Element |
Description |
|
External iteration |
Arrangement of the iteration steps of the external iteration in classes Note Double-click in the selected cell to edit the value. The maximum class limit is no subject to changes. |
|
Number of extra search iterations |
Number of search iterations for determination of further routes Note Double-click in the selected cell to edit the value. |
|
Create |
Via this button you can create additional classes. Notes A new class is automatically added after the currently selected class. For the new class, the class limit is automatically set to the middle between the two neighboring class limits. If the new class is added to the topmost limit, the maximum value is raised automatically. Double-click in the selected cell to edit the value. The maximum class limit is no subject to changes. |
|
Delete |
Via this button, you remove the currently selected row. |
3. Confirm with OK.
The number of search iterations has been set for the defined classes of external iterations.
Stochastic assignment: Preselection tab
In this tab, you define how to select the found routes.
|
Element |
Description |
|
A route is deleted if |
Here you define to which extent the current smoothed impedance and the travel time t0 of a route may exceed the values of the best route found. If all conditions are satisfied, the longer route is discarded. Here, the travel time t0 in the unloaded network is used. Note The smoothed impedance sums up from the impedances of all network objects traversed by the route. |
|
Apply test once when route is found for the first time |
If this option has been checked, only when finding the route for the first time, Visum checks whether it meets the criterion and should be dismissed. Afterwards, the route remains valid in all further iterations. This option is enabled by default. |
Stochastic assignment: Choice tab
In this tab, you define how to distribute the demand to the found routes.
|
Element |
Description |
|
Choice model |
From the drop down list, select the appropriate choice model. The model controls the distribution of the demand to the individual routes according to the route impedances. For each choice model, you can set certain parameters (Fundamentals: Distribution models in the assignment). |
|
Independence |
Via these options you decide how to calculate the independence of the routes. You select the attribute and calculation method to be used (Fundamentals: Similarity of routes and commonality factor). Define by t0 If this option has been checked, the similarity of two routes is calculated from the free-flow travel time. Here, the total routes are regarded. Define by length If this option has been checked, the similarity of two routes is calculated from the length. Here, the common section of the routes is regarded. Precise calculation according to Cascetta If this option has been selected, the commonality factor for the calculation of the independence of routes is calculated by means of the C Logit approach according to Cascetta. Instant approximate calculation according to Ben Akiva If this option has been checked, the Commonality factor, which enters the calculation of the independence, is calculated with a simpler approach according to Ben Akiva. Notes The calculation according to Cascetta requires a lot of calculation time due to the comparison of the routes in pairs; it is yet very precise. Ben Akiva saves calculation time yet impairs the accuracy of the calculation results. |
Stochastic assignment: Skims tab
You can calculate a skim matrix for the total impedance of the path, which includes the components of the impedance defined via the path. The skim matrix for the total path impedance is then calculated at the end of the stochastic assignment.
|
Element |
Description |
|
Skim 'Total path impedance' |
Considered OD pairs Here you can restrict the OD pairs to be evaluated to the active OD pairs, active origin zones, or pairs of active zones. Calculate only OD pairs with a demand > 0 If this option has been checked, only OD pairs with a demand > 0 are regarded. Weighting of paths Method for skim value determination from the skims of all paths resulting from assignment
Value of the path with the smallest value Maximum per skim Value of the path with the highest value
Mean value calculated from all paths
Mean value calculated from all paths and weighted by the particular path volumes Code of the skim Code of the skim for the total path impedance |
|
Calculate skim 'Total path impedance' for the following demand segments |
Here you define the demand segments for which the total path impedance is to be calculated. |
