Selecting the MSA method for travel times

With the Method of Successive Averages (MSA), you give each preceding iteration as much weight as the current iteration. This results in the arithmetic mean from all iterations. In this way, the influence of any further iteration becomes increasingly smaller.

The MSA method parameter depends on the cost file *.bew:

• If you select the option MSA (Method of Successive Averages) and there is no cost file *.bew saved yet by the dynamic assignment, the parameter of the MSA method will be set automatically by Vissim.
• If you have already performed a dynamic assignment and a cost file *.bew is stored, enter the number of iterations with which the file *.bew was created. If you enter a smaller value than the actual number of iterations, the subsequent iterations will be weighted higher. Enter a smaller value when the path evaluation shows that the measured travel times deviate significantly from the expected travel times (Showing data about paths of dynamic assignment in lists).

You can assign less importance to more distant measurements using exponential smoothing with smoothing factor for the travel times (Selecting exponential smoothing of the travel times).

1.  On the Traffic menu, click > Dynamic Assignment > Parameters (Attributes for the trip chain file, matrices, path file and cost file).

The Dynamic Assignment: Parameters window opens.

2.  Select the Cost tab.

3.  In the Smoothing method section, select MSA (Method of Successive Averages), so far.

4.  When a dynamic assignment was performed and thereby a cost file *.bew has been stored, enter the number of iterations with which the *.bew file was created in the field Iterations.

5.  Confirm with OK.

If the option Store costs is selected in the Files tab, the expected travel times are saved after every iteration for the next iteration in the Vissim cost file *.bew, from where they are entered into the path selection model.

After measurement of the new travel times, the smoothed travel time is computed for each edge as the weighted sum of the following:

• the old smoothed travel time from previous iterations
• the newly measured travel time from the current iteration

The new smoothed value represents the travel time that we expect in the next iteration.

Where:

N = user-defined value for number of existing iterations that shall be considered

K = index of the evaluation interval within the simulation time

n = index of the iteration

i = index of the edge

= measured (observed) travel time on edge i for interval k in iteration n

= smoothed travel time on edge i for interval k in iteration n

= variable smoothing factor from parameter N and the iteration index