Modeller Manual EMME Standard Toolbox Transit assignment (Tool Category) Extended transit assignment (Tool)
Extended transit assignment (Tool)
Section subtopics
Example 1: Optimal strategies
Example 2: Multiclass transit assignment
Example 3: Initial vs transfer boardings
Example 4: Must use Express mode
Example 5: Free transfers within same mode
Example 6: Discounted transfers between modes
Example 7: Maximum number of transfers
Example 8: Mixed-mode transit assignment using journey levels
Example 9: Distribution of flow at origins using fixed proportions
Example 10: Distribution of flow at origins based on time to destinations
Example 11: Distribution of flow between attractive lines also considering travel time
Example 12: Connector-to-connector paths prohibited
Tool syntax
Algorithm
Namespace
[Link].transit_assignment.extended_transit_assignment
Description
Performs a frequency-based transit assignment using optimal strategies and variants which can be used to model many aspects of transit
behavior. Optionally, saves resulting transit strategies for subsequent analysis. The Extended Transit Assignment is a multithreaded
procedure that will benefit from multiprocessing computer systems. See System requirements. The theoretical basis and technical details of
the Extended Transit Assignment are explained in the Algorithm section.
An extended transit assignment tool specification may contain the following details.
Assign more demand on existing transit volumes
By default, the new volumes (that is transit and auxiliary transit volumes, as well as boardings and alightings) replace the existing ones, but
they may also increment them. This makes it possible to obtain total volumes for several transit assignments performed with different
specifications, for example different sets of active modes or different perception factors. The strategies from each assignment can be saved as
a separate class and then accessed for post-assignment analysis (see Save strategies). The status of the scenario will reflect all the classes
assigned. Provided that they are all given class names (see Save strategies), the sequence of assignments will be listed by name.
Network and demand
Modes: The set of transit and auxiliary transit modes which define the subnetwork for which the assignment is to be performed. If a link
allows for more than one of the auxiliary transit modes specified, only the fastest one will actually be active.
Demand: A matrix containing, for each O-D pair, the number of travellers.
While it is possible to run a transit assignment on a subnetwork of only auxiliary transit modes, at least one transit line must be defined in the network. In
this case, the mode of this dummy transit line should not be specified in the assignment.
Times and costs
Waiting time
The waiting time is computed by using the headway fraction and the combined effective headways of the attractive lines, and then
multiplied by the product of the perception factor and the spread factor.
Effective headways, to specify the headway actually perceived by the users of a transit line. May be the actual line headway (hdw),
or a perceived headway which is line or segment specific. Headways must be between 0.01 and 999.99 minutes.
The effective headways are used in the assignment, for computing the waiting times and the transit volumes.
The actual headways are used for computing statistics such as the number of buses required to operate a line, the operating costs and the energy
consumption.
Headway fraction. May be constant or node-specific. Headway fractions must be greater than 0 and less than or equal to 1.00. A
value of 0.5 for this parameter corresponds to a regularly spaced service which results in a waiting time of half the combined
headway of the attractive lines. A higher value can be used to model situations in which the service spacing is irregular (for
example, vehicle bunching), whereas a lower value may be used if the arrival of the passengers at the boarding node is timed to
coincide with vehicle arrival (for example, passengers know the time table or transfers at node are coordinated).
Spread factor, to adjust the waiting time to give the travellers fewer or more alternatives at a stop, given the same perception
factor. It may be used as a model calibration parameter.
Different combinations of waiting time perception factor and spread factor that result in the same product, yield the same results;
this makes it possible to use the same waiting time perception factor as in the mode choice model, while achieving the desired
spread level. Given the same waiting time perception factor, the higher the spread factor, the more travellers are inclined to
consider several attractive lines at a given stop; small spread factor values result in strategies which are closer to a single path.
By default, the spread factor is assumed to be 1. The value for the spread factor must be greater than or equal to 0.
Perception factor, to quantify the perception of waiting. For example a perception factor of 2.5 for the waiting time means that
travellers perceive 1 minute spent waiting as equivalent to 2.5 minutes. May be constant or node-specific values.
Note that waiting time perception factor * spread factor must also be greater than or equal to 0.
Boarding time
The boarding time is a penalty associated with every boarding, be it an initial or a transfer boarding. This penalty might represent the
time spent in a station (corridors, stairs, etc.).
Specify a global penalty to apply the same value across the whole network, or specify a node, line or segment attribute for element-
specific boarding times. Note that for segments, a node, link, or transit line attribute is also accepted, but doing so will slow down the
assignment.
When penalties on more than one type of element are specified, for example, at nodes and on lines, the sum of the values (for line L at
node N) is used.
The corresponding perception factor may be a constant or an attribute of corresponding type. For segments, the perception factor may
be also be a node, link or transit line attribute.
The boarding time penalty and its perception factor must be greater than or equal to 0.00.
Boarding cost
The boarding cost is also a penalty that is associated with every boarding, be it an initial or a transfer boarding. This penalty might
represent the fare (converted to generalized minutes) paid upon boarding.
The boarding cost is specified in the same way as the boarding time. The perceived boarding time and perceived boarding cost are
added together and applied to every boarding.
In-vehicle time
The in-vehicle time perception factor will multiply the transit times (timtr) resulting from the evaluation of the transit time functions.
The in-vehicle time perception factor may be constant or segment specific (a segment, transit line, link or node attribute can be
specified). Its value(s) must be greater than or equal to 0.00.
If the transit time function of a segment cannot be evaluated, the line speed is used to compute the transit time for this segment. This happens when:
The transit time function uses the auto time (timau) and the auto time is not available on the link, either because the auto mode is not defined on
the link or the results of a traffic assignment are not available.
The transit time function to which a segment refers is not defined.
In-vehicle cost
An optional in-vehicle cost attribute can be specified, which will be multiplied by the specified in-vehicle cost perception factor, and
added to the perceived in-vehicle time.
The in-vehicle cost perception factor may be constant or segment specific (a segment, transit line, link or node attribute can be
specified). Its value(s) must be greater than or equal to 0.00.
Auxiliary transit time and cost
Auxiliary transit time and cost can be expressed in two ways: Globally or By mode.
Global auxiliary transit time and cost
Global auxiliary transit time
The auxiliary transit time on links will be computed based on the speed factor of the auxiliary transit modes used on the link
according to the modes definition (in 1. Network and demand).
The auxiliary transit time is multiplied by the time perception factor, which may be constant or link-specific (a link or node
attribute may be specified). Its value(s) must be greater than or equal to 0.00.
Global auxiliary transit cost
An optional auxiliary transit cost link attribute can be used, which will be multiplied by the specified auxiliary transit cost
perception factor and added to the perceived auxiliary transit time.
The auxiliary transit cost perception factor may be constant or link-specific (a link or node attribute may be specified). Its
value(s) must be greater than or equal to 0.00. If left empty, a zero auxiliary transit cost will be assumed.
If more than one active auxiliary transit mode is defined on a link, only the fastest mode is used. In the case of a tie, the first
mode appearing in a mode table in Desktop will be considered. For example, in the mode table below, the pedestrian mode
(mode=p) would be considered on links where both auxiliary transit modes are defined.
By mode auxiliary transit time and cost
The By mode auxiliary transit time option allows to define the time, cost and perception factors for each auxiliary transit mode
specified in the 1. Network and demand panel. These values, specified as link extra attributes or constants (where applicable) are
used instead of the mode speed factor in the assignment.
The mode definition remains unchanged in the database; it is just that the mode speed factor that is ignored during the assignment.
Values must be specified for all auxiliary transit modes listed, otherwise the tool will raise an error.
When using the By mode option the following four values may be specified:
Time: Required for all modes in the list.
Time perception factor: Optional. If no time perception factor is specified, a default value of 1 will be used.
Cost: Optional. If no cost attribute is specified, a default value of 0 is used.
Cost perception factor: Optional, if a mode-specific cost is specified. If no cost perception factor is used, a default value of 1
is used.
The auxiliary transit generalized cost is obtained by adding the perceived auxiliary transit cost to the perceived auxiliary transit time.
When no journey levels are defined, if two auxiliary transit modes will yield the same auxiliary generalized cost, the fastest mode will be used. In
case of a tie, the first auxiliary transit mode appearing in the mode table will be used.
When journey levels are defined, if two auxiliary transit modes with the same auxiliary transit generalized cost share the same journey level
transition, they become ambiguous and the first mode listed in the mode table will be used. If the link flows per mode are requested for a mode
that has been detected as ambiguous, a specific warning will be generated. In this case, the obtained flow for an ambiguous mode will contain the
cumulative value from all modes sharing the same cost.
Journey levels
Journey levels provide a facility to vary the generalized costs experienced by travellers along their journey according to the particular
sequence(s) of transit and auxiliary transit modes that they use.
Journey levels are user-defined, and allow explicit modelling of:
Transit fare integration schemes, for example free/discounted transfers or fare rebates on combined mode trips
Reduced waiting time on initial boardings, for example, from reliable transit itinerary information
Must-use or forced transit mode, in which travellers are forced to use a particular transit mode, or combination or sequence of transit
modes. This includes mixed-mode transit journeys in which public transit may be used together with private modes but only in certain
specific sequences, for instance first- or last-mile drive access/egress trip legs (including park-and-ride), ride-hailing/taxi trip legs
from/to designated pickup/drop-off locations or intermediate trip legs using micromobility options.
All travellers start in journey level 0. Additional journey levels may be created to represent the different generalized costs experienced in
terms of waiting time, boarding time or boarding cost as travellers use transit modes through their journey. When travellers board a transit or
auxiliary transit mode, a transition rule in the traveller’s current journey level determines whether they will continue to experience the
generalized costs at the current journey level, or different generalized costs at another journey level.
Consider the example of the small transit network below. A traveller walks to his first boarding point and has the choice of taking one of
three transit modes. Once the traveller pays the fare and boards any transit mode, he goes to level 1. If he does not board a transit mode, he
remains at level 0. From level 1 he can transfer to any other transit mode for an additional cost (full fare, discounted fare or no cost).
Transfer boardings may be penalized more than initial boardings since travellers may be aware of the schedule of the transit lines at the first
boarding point, thereby reducing their waiting time for the initial boarding.
In this first strategy, the traveller incurs an initial boarding cost at A, at level 0, and a transfer boarding cost at B' and other transfer points, at
level 1.
In this other strategy, the traveller incurs an initial boarding cost at B, at level 0, and a transfer boarding cost at further transfer points, at
level 1.
If the traveller does not board a transit mode, he can still walk to the destination, thus never leaving level 0.
For each journey level, it is possible to specify:
Transition rules :
transition rules determining at which journey level travellers will be after boarding a given transit or auxiliary transit mode
whether destinations are reachable at this level
Times and costs:
waiting time parameters for travellers boarding at this level (effective headways, headway fraction, perception factor and spread
factor)
boarding time and cost parameters for travellers boarding at this level (global or node, line, and/or segment penalty and
perception factor)
When a checkbox is specified for either waiting time, boarding time or boarding cost in the Times and costs tab of the Journey levels
panel, the values displayed for each level will override the corresponding values specified in the [Link] and costs panel. Changing
values in the 2. Times and costs panel will not be propagated to the Times and cost tab of the Journey levels panel tab after the journey
levels have been specified.
Flow distribution at origins
In the optimal strategy transit assignment, all the flow of an O-D pair uses one outgoing connector: the one with the best transit time to the
destination. In the logit strategies transit assignment, it is possible to distribute the flow between several connectors using a logit distribution.
Consider the example of the small transit network below. Travellers going from P to Q have three options: walk to B to take the metro, walk
to node A to take the bus or walk to node C to take the streetcar. The travel time ( t) and headway for each transit line are specified in the
illustration below. The optimal strategy would assign all travellers to the fastest mode, which in this case, is the metro.
If we specify a logit choice on connectors, we would obtain a more even distribution on the services offered. Using a logit choice with a scale
parameter of 0.1 yields the following flows.
Choices at origins: Specify the choices for travellers at an origin:
optimal strategy: no logit distribution at any origin; all travellers leave their origin by the connector yielding the best transit time
to their destination
logit distribution: the logit distribution is applied at the specified choice points.
choice points: may be all origins or only those specified via a node attribute. If using a node attribute, use 0 to indicate that
the optimal strategy should be applied at that node, and use 1 to indicate that a logit distribution should be applied.
choice set: at specified choice points, logit may be applied to all connectors or only to efficient ones; an efficient connector
is a connector that brings the traveller closer to the destination.
Specifying a node attribute makes it possible to use different choice sets at different origins: use 1 to indicate that the logit is
applied to all connectors, and use −1 to apply logit to efficient connectors only.
Logit function parameters:
Scale: parameter used in the computation of the likelihoods on which the proportions are based. It must be greater or equal to 0.
A value of 0 results in equal proportions for all connectors in the choice set. The larger the value, the higher the proportion for the
best connector.
Truncation : parameter used to drop connectors with a proportion considered too small. The proportions are first computed for
each connector in the choice set. Then connectors with a proportion inferior to the truncation/cutoff parameter are dropped, until
the proportion of each remaining connector is sufficiently large.
The truncation parameter must be between 0 and 1. With a value of 1, only the best connector is retained.
Both the scale parameter and the truncation parameter may be constant or node specific.
Override flow connectors with fixed proportion
The computed connector proportions can be overridden by fixed proportions, provided in a user-defined link attribute. The proportions
must be between 0 and 1, and the sum of the proportions for all the connectors outgoing of a given origin must be 1 (with a tolerance
of 0.0001). Proportions can be provided only for a subset of origins; for the connectors outgoing from other origins, the link attribute
must be set to −1. Origins where the fixed proportions attribute value is 0 on all outgoing connectors will be excluded from the fixed
proportions subset; this is equivalent to specifying a value of −1 on all outgoing connectors for these origins.
When, for a destination, there is no path from a connector with a non-zero user-defined proportion, the demand for this O-D pair is not
assigned at all. The report then includes a warning for the problematic origins.
Flow distribution at regular nodes with auxiliary transit choices
In the optimal strategy transit assignment, all travellers at a node wait there for a vehicle of an attractive line or leave by an auxiliary transit
link, depending on the option that yields the best transit time to the destination. In the logit transit assignment, travellers at a node consider
the following options:
wait at the node for a vehicle of an attractive line
leave the node by auxiliary transit; the choice set considered can be:
only the best link
all the efficient links, that is links that bring the traveller closer to the destination (the transit time to destination is less
from the J-node than from the I-node)
all the possible links, except those links that would introduce a loop in the strategy.
Consider a slightly different network than the one above. Now, only one connector leaves origin P. Once at node A, the choices are now: wait
at A for the next available bus; walk to B to take the metro; or walk to C to take the streetcar. The travel time on each transit line is the same
as in the network above; the walk times are shown in the illustration below.
The optimal strategy would assign all flow to the walk + metro option, since that is the option that yields the best transit time to destination
Q.
Applying a logit choice at node A on all efficient links with a scale parameter of 0.1, the flow is distributed between ride and walk-and-ride
options as shown below.
Choices at regular nodes: Specify the choices for travellers at a node:
Optimal strategy: no logit distribution at any node
Logit distribution : a logit distribution is applied at the specified choice points.
choice points: may be all origins or only those specified via a node attribute. If using a node attribute, use 0 to indicate that
the optimal strategy should be applied at that node, and use 1 to indicate that a logit distribution should be applied.
At nodes with logit distribution between walk and wait, it is possible to also split travellers between alighting and
staying on-board. However, this split will actually occur only if:
the line on which the travellers are is also attractive at the node
and
it is also possible to leave the node by auxiliary transit.
If both these conditions are met, a proportion p_alight of travellers alight at that stop; they must leave the node by
auxiliary transit. The remaining travellers stay on-board. At the first node where the line is not attractive, all remaining
travellers alight.
The proportions for alighting/staying on-board at a node are computed based on the corresponding impedance to the
destination. The scale parameter and the truncation/cutoff parameter of the node apply.
The logit distribution cannot be used to distribute the flow at nodes with auxiliary transit choices when the scenario contains lines with
intermediate layovers.
Auxiliary transit choice set: At the specified choice points, the logit distribution may be applied to the best link, to all the
efficient links, or to all possible links, in addition to the wait option, if available. Specifying a node attribute makes it possible
to use different choice sets at different nodes: use 1 to indicate that the logit is applied to the best link, use −1 to apply logit
to efficient links, and −2 for all possible links.
Logit function parameters:
Scale: parameter used in the computation of the likelihoods on which the proportions are based. It must be greater or equal to 0.
A value of 0 results in equal proportions for all options that can be used. The larger the value, the higher the proportion for the
best option.
Truncation : parameter used to drop options with a proportion considered too small. The proportions are first computed for the
wait option and for each link in the choice set; options with a proportion inferior to the truncation parameter are dropped, until
the proportion of each remaining option is sufficiently large. The truncation parameter must be between 0 and 1. With a value of
1, only the best option is retained.
Both the scale parameter and the truncation parameter may be constant or node-specific.
The use of the logit model also results in smoother responses to changes of the transit services. Suppose that the travel time by the metro
mode changes. The changes in the flows for each mode are indicated below for the logit choice at node A. As the metro flow decreases the
flow on both the bus and the tram increase.
The optimal strategy chooses the best mode at node A. As the time of the metro mode increases the flow change is more abrupt from the
metro to the tram when the tram time becomes less than the metro time.
Flow distribution between lines
By default, the flow at a transit stop is distributed based on frequency only. It is also possible to distribute the flow based on frequency and
impedance to destination. In this case, fast but infrequent transit lines are favored. Furthermore, the changes in line travel times result in
smoother flow changes.
The impedance to destination may be considered: never (which is the same as distributing flow by frequency only), at all nodes, or at
selected nodes via a node attribute that is set to a non-zero value for nodes where both frequency and impedance to destination are to be
considered, and to 0 for nodes where only frequency is to be considered.
Connector to connector path prohibition
Connector-to-connector paths, that is, paths consisting only of two connectors, may occur when either the network has large zones, or when
the network is dense. In either case, the total impedance to the adjacent zone is greater if the traveller would take a transit or auxiliary transit
service rather than only taking the two connectors.
By default, connector-to-connector paths are allowed. It is possible to prohibit some or all of these paths. For prohibited paths, the
corresponding demand may be reassigned to an alternate path (the best one using a different outgoing connector from the origin); otherwise
it is unassigned.
Note that there is an alternative method to prohibit all-walk paths that does not result in unassigned demand. It is recommended instead to use journey levels
with the destination reachable option; this prohibits explicitly the all-walk option, while allowing the same connector(s) to be used for a different transit strategy.
Connector-to-connector paths may be prohibited at all nodes, or only at certain zones and/or regular nodes (specified via a node attribute set
to a non-zero value for nodes where connector-to-connector paths are prohibited, and to 0 for nodes where they are allowed).
Note that paths do not pass through centroids.
Circular lines
Lines that start and end at the same node and that have only one layover can be considered as circular lines. In the extended transit
assignment, there are three options to handle the behavior of passengers at the end of such a line:
Force passengers to alight at terminus (standard): All passengers are forced to alight at the end of the line. No passenger incurs the
layover time.
Allow passengers to stay on-board: Passengers are allowed to stay on-board on all circular transit lines. The layover is added to their
in-vehicle travel time.
Allow passengers to stay on-board for certain lines: Same as above, but only for those circular transit lines identified in a line extra
attribute or user data item.
Results
The total impedance, that is the sum of perceived time and cost components, may be stored in a matrix. It is also possible to recompute it
later, using the Matrix results (Tool) , if the strategies have been saved. In both cases,
Results for each O-D pair are obtained by specifying a full matrix (mfN).
Destination averages, weighted by the demand matrix, are obtained by specifying a destination matrix (mdN).
A weighted global average is obtained by specifying a scalar matrix (msN).
If an origin matrix is specified (moN), each cell of this matrix will contain the global average, not the origin average. (This is due to the
fact that the results are only available on a destination level; the automatic expansion is thus used to duplicate the global average into
all origin cells.)
When saving results in a matrix other than a full matrix, the sum of the demand used to compute the average result does not include the intrazonal demand.
Auxiliary transit volumes may be saved to link attributes for each auxiliary transit mode specified in the assignment. By default, only the
fastest auxiliary transit mode will be used by travellers on a link. However, with the use of journey levels, it is possible to specify that the
transit modes available on a link vary over the transit journey. In this case it may be useful to analyze the specific auxiliary transit mode
volumes used.
Performance settings
The number of processors determines how many processors will be used for the extended transit assignment. It is possible to use as few or
as many processors as desired, for instance to leave a processor for other concurrent processing, or to dedicate all processors for maximum
performance. The drop-down menu gives the possible choices for the number of processors for the system on which Modeller is currently
running. Select the Maximum or Maximum − 1 option to always run the assignment with the maximum or one less than the maximum
number of processors available on any system.
The term processor is used informally in that multithreaded systems (that is, Intel hyper-threading) will add virtual cores to the processor count, in addition to
processors from multiprocessor and multicore systems.
The parallelized procedure may be more RAM intensive, since it requires keeping in memory all the information to be written in the strategy file.
Setting number of processors to 1 processor, non-parallelized procedure (or when scripting, specifying 0) will use 1 CPU, but will write the strategies to
the strategy file as they are computed.
The memory consumption depends on the network size and the assignment setup. If a serious slow-down occurs during the assignment, it
might mean that not enough physical RAM is available. When not enough memory is available to create new threads or to keep the whole
strategy in RAM, OpenPaths EMME exits with a STOP message.
Save strategies
The strategies computed during the assignment can be saved in a file named according to the scenario number (STRAT_sN). This strategy
file is saved in the Database folder, when the option is selected.
If more demand is assigned to existing transit volumes to either increment the transit volumes on the network or to run an assignment with
different specifications, it is possible to save the strategies from each assignment in a separate class. To do this, a class name must be
specified. Any name can be used but it must not contain the following characters: ' " < >. The strategies will then be written to a folder
named according to the scenario number (STRATS_sN).
If no class name is specified and the save strategies option is checked, a default strategy file will be written and any previous strategy files will
be removed.
Post-assignment analysis using the tools in the Extended transit analysis (Tool Category) is possible only if the strategy file (or class of
strategy files) is available.
Network results
After the assignment has been performed, the following results are always available:
Transit times on segments ( timtr), in minutes. The transit time is the sum of the travel time (given by the transit time function) plus
the dwell time at the J-node.
Transit boardings ( board) and alightings (alight) at I-node of segments.
Transit volumes ( voltr) on segments.
Initial boardings (inboa) and final alightings (fiali) at nodes.
Auxiliary transit volumes on links (volax). Note that this is the total auxiliary transit volume over all auxiliary transit modes permitted
on the link. To analyze the volumes per auxiliary transit mode see Results above.
All volumes, boardings and alightings correspond to the assignment period (see Assignment Period).
Figure 8. Extended transit assignment inputs and outputs
Tool report
When the Extended transit assignment tool is run, a report and a transit assignment summary are written to the Logbook after assignment
execution. If the Tool report path project setting is set, the tool report is also written to a file in the specified location.
The report begins with a summary of the specification for the assignment and summary statistics for the network. The main part of the
report is a summary for each destination, including:
A summary of the demand:
the total demand from all origins to the current destination
the part of the total demand that has been assigned on the network
the intrazonal demand, that is, the demand for trips originating from the current destination
the part of the total demand for which no path exists, and that was consequently not assigned on the network.
Boarding information:
the number of trips that used auxiliary transit modes only
Note that the number of trips using auxiliary transit only is not computed when using logit distribution and appears as "n/a" in the assignment
summary. However, the matrix of the auxiliary transit only trips can be obtained from a strategy-based analysis:
Boarding component: line attribute set to 1 for all lines
Sub-path combination operator: +
Sub-strategy combination operator: Min
Sub-strategies to retain: From combination operator
Selection threshold: 0 (Lower), 0 (Upper)
Save the resulting strategy values, which will contain the number of aux. transit only trips for each O-D pair.
the total number of boardings
the average number of transit lines used per passenger.
Trip time information:
the total passenger hours, which is the sum, over all the origins, of the demand multiplied by the total impedance
the mean impedance, which is obtained by dividing the total passenger hours by the total assigned demand. Note that the total
assigned demand used to compute the mean impedance does not include the intrazonal demand.
At the end of the report, global statistics are given. They are the sum over all the destinations of the corresponding statistics (except for the
mean impedance, which is the mean for all the destinations).
The transit assignment summary makes it easier to analyze the information contained in the report. It begins with a summary for all the
zones, giving the global statistics described above. It then displays the information from the report in a graph. Each column of the report can
be drawn alternately on the graph by selecting the desired column. See also Zooming in Logbook charts.
Copyright © 1983-2024 Bentley Systems, Incorporated. All rights reserved.
Frequency optimization (Tool) Example 1: Optimal strategies