Additional Options for Transit Assignment

Ever since the introduction of the additional options for the auto assignment, transit modelers have been longing to get access to a similar feature for the transit assignment. While it was always possible -with some work-arounds- to find the O-D pairs which are using or not using a certain facility, there was no way to obtain the precise part of the demand, nor to reliably compute additional attribute matrices, such as a true transit distance matrix.

Transit assignment users will be happy that Release 7 does provide this long awaited new feature. Given the very different nature of auto and transit assignment, it is not possible, nor desirable, to implement the additional options for transit in exactly the same way as it is done in the auto assignment. So, before discussing the actual implementation, let us look again at the main differences between the two assignment models:

From the above comparison, it follows that the implementation of the additional options for transit has to be quite different from the one for the auto assignment. Assuming that the reader is already familiar with the general concept of the additional options for the auto assignment, here is a summary of the main differences for the transit implementation of the similar functionality:

• Since a strategy is composed of boarding, riding, alighting and walking activities, not just one but several attribute vectors are needed to characterize the set of strategies which are of interest for the analysis. Three additional segment attributes, boarding, invehicle and alighting attribute, as well as an additional link attribute, the auxiliary transit attribute, can be used to characterize the strategies to be retained.
• Since a strategy is not a simple linear structure, but splits into sub-strategies at each boarding point, two distinct operators are required to compute, for a given strategy and the specified additional attribute values, the corresponding strategy attribute, which is then used to determine if the strategy is to be included or excluded from the analysis. The path operator is used to combine attribute values along the linear parts of the sub-strategies, whereas the sub-strategy combination operator is used for combining the attribute values of the sub-strategies at the boarding points.
• The transit assignment is a fixed cost assignment in which the route choice does not depend on the resulting volumes, as is the case in the auto assignment. Thus, there is no need to implement the additional options assignment as a slave assignment in parallel with the standard assignment. Instead, the additional options assignment is done instead of the standard assignment and the assignment results are stored in the standard result attributes voltr, volax, board, inboai and fialii.
• A typical use of the additional options is to compute the partial volumes of only those trips which satisfy a certain condition, e.g. use a given line or board at a given node. Since the trips are probabilistically distributed onto the sub-strategies, not all trips assigned to the same strategy satisfy the same condition. In addition to the possibility of retaining or rejecting entire strategies, it is therefore necessary to provide means of retaining trips at a sub-strategy level. Most applications of this type can be solved by retaining the trips assigned to the minimum or maximum attribute sub-strategies, i.e. those sub-strategies which are minimal or maximal with respect to the specified additional attributes.

It is, of course, not possible here to go into all the technical details of the implementation of the additional options for transit. For in-depth technical information, please refer to the corresponding pages in the Release Notes and the EMME/2 User's Manual. Instead, let us look at some interesting applications which become possible with this new development:

• Computation of distance matrices: The average transit distance matrix can be obtained easily by specifying length as additional invehicle and auxiliary transit attribute, and using + as path operator and the default sub-strategy combination operator average. Starting from the destination, this will compute the average transit distance of each O-D pair by summing the link lengths along the linear parts of the optimal strategy and by averaging the distances of all outgoing sub-strategies at each boarding node, using weights which correspond to the respective probabilities. Many variants are possible, e.g. the computation of invehicle distances, or mode, line, or district specific distances. This is done simply by modifying the additional attributes accordingly.
• Computation of fare matrices: Fare matrices for any kind of additive fare structure can be implemented by defining the additional attributes to contain the corresponding cost component, usually at the level of boarding and invehicle segments. E.g. if the fare is based on the number of fare zone boundaries crossed, it suffices to use an invehicle attribute in which segments on fare zone boundaries are coded as 1, all others as 0.
• Select line assignment: Similar to the select link analysis for the auto assignment, the select line analysis is used to retain only volumes of those trips which somewhere along their itinerary are using a given transit line. This type of analysis can be achieved by creating an extra line attribute which contains 1 for the line to be analized, 0 for all other lines. This attribute is used as additional boarding attribute (invehicle or alighting attribute would also work in this case), and the operator .max. is used for both path and substrategy combination operator. The threshold values (1,1) ensure that only strategies using the line are considered active, and retaining not the complete strategies but only the maximum attribute substrategies ensures that only those trips which are actually assigned to the line in question are retained.
• Select mode assignment: Note that the simple method outlined above for the select line analysis is based on the hypothesis that the selected line is boarded at most once by each passenger. It is therefore not applicable to analyzing all trips on a set of lines or a particular mode. In this case, it is easier to retain the complementary trips, i.e. those trips which do not use any of the specified lines, by still using .max. as path operator, but retaining the minimum attribute sub-strategies and using the threshold values (0,0).

To carry out an additional options assignment, module 5.11 is used to prepare the scenario for a transit assignment as usual. At the very end of the standard assignment preparation dialog, the following question has been added:

   Perform additional options assignment?

When answering ``no'' to the above question, module 5.11 is left and the scenario is ready to be assigned. Thus, only a minimal change is needed to update existing transit assignment macros to be compatible with this new development: it suffices to add a ``no''-line just after the line containing the boarding time weight (which is usually just before the call to module 5.31).

If the additional options are activated by answering ``yes'' to the above question, a dialog is generated, in which the user is first prompted to specify the additional boarding, invehicle, alighting and auxiliary transit attributes which will be used to characterize the strategies. Only those attributes which are pertinent for the desired analysis have to be specified, non-specified attributes defaulting to zero. Also, full attribute inheritance is implemented for the additional attributes, e.g. if a transit line attribute is specified as the boarding segment attribute, the value for each line will automatically be copied to all its transit segments. The dialog proceeds then to prompt the user for the part of the strategies to be retained during the assignment, the path and sub-strategy combination operators and the threshold values for determining active strategies. Finally, the user can choose to save two result matrices: the additional strategy attribute matrix contains the values obtained by applying the path and sub-strategy combination operator to the additional attribute values along the strategy, and the active transit demand matrix contains that part of the demand which satisfies the given condition, i.e. the demand which corresponds to the retained transit volumes.

The following example illustrates the new dialog in module 5.11 for a select line analysis. The plot below shows the resulting additional volumes, i.e. the volumes of all trips using the selected line.

Perform additional options assignment? yes

Enter: Additional boarding attribute= @selin
T: @selin  select line 24                            94-05-22 08:02
Enter: Additional invehicle attribute= 
Enter: Additional alighting attribute= 
Enter: Additional auxiliary transit attribute= 

Select: Retained part of strategies
        1= Complete strategies
        2= Minimum attribute sub-strategies
        3= Maximum attribute sub-strategies
          3

Enter: Path operator(+)= .max.

Enter: Lower, upper threshold for active strategies= 1 1

Matrix to hold additional strategy attributes (optional)
Enter: Matrix( mf )= 

Matrix to hold active transit demand (optional)
Enter: Matrix( mf )= line24
mf04: line24     partial demand using line 24             (94-04-30 11:43)
Change header information? no