The resident model is actually a series of models that attempt to capture multiple aspects of household, person and tour-level travel making decisions. When applied these components essentially take the place of trip generation, trip distribution and mode split that are the more familiar steps of a 4-step model. However because this is an activity-based model, the components cannot be grouped the same way since trip generation is done on a person-by-person basis with trip distribution and mode split integrated along the way.
The population synthesis procedure is designed to create a list of households in each TAZ with all necessary details regarding the household and person variables used in the travel models and according to specific zonal characteristics; namely, average number of workers in a household per zone, average household size and number of households per income group. The basic idea is to set up a 3-dimensional table for each zone with
The cells in the table are given seed values based on what is given by the Census Transportation Planning Package (CTPP) data for the census tract that contains the zone. A procedure called iterative proportional fitting (IPF) or matrix-balancing is then used to adjust the seed values to match the marginal distributions in all dimensions. Once the table has been balanced, each cell is multiplied by the total number of households to give the total number of households per category (size, worker, income combination). Household records from the Public Use Micro-Sample Area (PUMA) that encompasses the zone are randomly drawn that match the given category until all of the households in the table have been accounted for. When the procedure is done for every zone, then the synthetic household and person files are written and saved for later use.
The number of autos available to a household is an important variable for explaining household travel behavior. It is included in such subsequent models as tour generation, mode choice, destination choice, and stop frequency/location choice. In the Tahoe AB Model, auto ownership is considered a household-attribute variable; thus, the auto-ownership choice model employs only household and zonal characteristics and is applied before any travel-related model. There are five naturally ordered alternatives;
The model is a multinomial logit model and predicts each household to have one of these auto ownership levels.
In the Tahoe AB Model each person in the model area gets to explicitly choose whether to leave the house on the model day and if so whether to go to work and/or school (mandatory tours) or do some kind of non-mandatory travel (joint and/or individual non-mandatory tours). This decision is modeled by the Daily Activity Pattern (DAP) model. A person’s daily activity pattern (DAP) is classified by one of three main types:
The DAP decision is also modeled using a multinomial logit model. The DAP model used in Tahoe AB Model is a coordinated DAP model. This means that the decisions of different individuals in a households are correlated/coordinated. The DAP model has the following alternatives in the model.
Segmented by Person Type
If a person in a household chooses a daily activity pattern with a work or school component, then that person is said to be making a mandatory tour. The mandatory tour destination (D), time-of-day (T), and mode choice model (M) (DTM) determines where that tour will go (the destination), when the tour will happen (the time-of-day), and what mode the person will use to travel during the tour (car, bus, etc). If the daily activity pattern chosen by the individual includes both school and work, then the school tour is processed first, followed by the work tour.
The destination choice model is a multinomial logit model in which each potential destination zone is an alternative. The probability of each zone being chosen is calculated from a utility function, where the utility consists of variables such as distance, income level, and area type. To provide a measure of a zone’s attractiveness based on tour-specific characteristics, a size term is included in the utility expression. Also included in the utility expression is the logsum from the mode choice model, which provides accessibility indices for a destination zone - the higher the logsum, the more accessible (by auto, transit, walking) a zone is.
In the mandatory work destination choice model, the size term (attractiveness measure) of a zone is based on the employment in that zone. However, the destination choice model places no constraints on how many people can choose a given zone for their work destination. Thus, it is possible that more people choose a particular zone as their work destination then there are employment spots. This overfilling of employment is unrealistic and, especially since the target employment distribution among zones is known, should be addressed.
The solution for the overfilling of zonal employment in the model is to use shadow pricing. In this scheme, the mandatory work destination choice model is run several times. If the number of work tours choosing a zone as the destination exceeds the employment in that zone, then a negative penalty (shadow price) is added to the utility of that zone. Conversely, if a zone’s employment is underfilled, a positive shadow price is added to the utility. After a number of iterations, the result is that the work destination choice distribution among zones more closely matches that of the actual employment distribution.
The time-of-day sub-model is a multinomial logit model in which start/stop hour pairs make up the alternatives. The earliest allowed start/stop time is 5:00 am (corresponding to the 5:00-6:00 hour), and the latest allowed is midnight (corresponding to the 12:00am-1:00am hour). As far as skim periods are concerned, the following definitions are used:
| Skim Period | Start Time | End Time | Duration |
|---|---|---|---|
| AM Peak (AM) | 7:00 AM | 10:00 AM | 3 hours |
| Midday (MD) | 10:00 AM | 4:00 PM | 6 hours |
| PM Peak (PM) | 4:00 PM | 7:00 PM | 3 hours |
| Late Night (LN) | 7:00 PM | 7:00 AM | 12 hours |
The mode choice model is a multinomial logit model in which each mode is an alternative. For the mandatory tours, the following alternatives are available:
The primary component of the model is travel time, which uses the same coefficient across all modes. For the modes that have costs associated with them (transit has fares, auto modes have operating costs), a value of time factor was estimated; this factor converts dollar costs into time costs, for which a utility can be calculated using the travel time coefficient.
Mode Choice Model * work tab used for work tours * school tab used for school tours
The joint travel types modeled explicitly in the Tahoe AB Model are limited to fully joint tours generated by shared non-mandatory activity of several household members. Joint Tour sub-models can be grouped into two aggregate categories:
These components are described in detail in the next sub-sections.
Generation of joint travel is basically an entire-household function, thus the tour-frequency model comes first and is applied at the household level. In order to link joint travel to the persons in the household, two additional models - travel party composition and person participation - are then applied.
This model is applied to each household. Following are the alternatives of this model:
This sub-model is applied to each of the joint tours generated by the joint tour frequency sub-model. The alternatives of this sub-model are the following:
It is applied to each joint tour listed in combination with each household member suitable for the travel party and not staying at home. The model has only two alternatives: the person participates in the joint tour or the person does not participate in the joint tour. This sub-model has different eligibility rules and iterations to make sure that the model comes up with a prediction that is consistent with the travel party composition sub-model.
At the end of the above three joint tour generation sub-models, the number of joint tours undertaken by each household and the composition (persons) of each of them is known. The model still needs to predict the characteristics of these joint tours (destination, TOD and mode). This is done in the next model component, the Joint Tour DTM Component.
If a household chooses to make a joint tour, the joint tour destination, time-of-day, and mode choice model (DTM) determines where that tour will go (the destination), when the tour will happen (the time-of-day), and how the tour participants will travel during the tour (the mode). When the model is applied, each tour party making a joint tour is treated as a separate and independent decision making unit.
The DTM models for the joint tours are very similar to the mandatory tour’s DTM. One important difference is that joint tour destination choice models do not have demand constraining/shadow pricing. This is because, unlike mandatory tours, joint tours (which are non-mandatory) do not have destination constraints.
Individual non-mandatory tour models are also composed of several sub-models and can be grouped into two aggregate categories:
Each of these two components will be described next in detail.
The individual tour generation model for non-mandatory activity includes 3 choice sub-models applied successively:
Individual tours generated by allocated maintenance activities are modeled first for each person conditional upon the chosen daily pattern and participation in joint household tours. Since these activities are generated by the entire household and then allocated to particular members, it is important to follow an underlying intra-household allocation process.
Individual tours for personal discretionary activities are modeled next because they normally have a lower priority in scheduling. Intra-household linkage is less important at this stage. Person availability in terms of time window left after scheduling the mandatory activities, joint activities, and allocated activities becomes the most crucial determinant.
Work-based sub-tours are modeled last. They are relevant only for those persons who implement at least one work tour. These underlying activities are mostly individual (business-related and eating-out purposes), but may include some household maintenance functions as well that are linked to the person and entire-household maintenance tasks.
If a person chooses to make an individual non-mandatory tour, the individual non-mandatory tour destination, time-of-day, and mode choice model (DTM) determines where that tour will go (the destination), when the tour will happen (the time-of-day), and how the tour participants will travel during the tour (the mode). The DTM models for the individual non-mandatory tours are very similar to the joint tour’s DTM.
In any given tour up to one outbound and one inbound stop is allowed. An outbound stop is one that occurs during the trip to the primary destination, whereas an inbound stop is one that occurs on the way back to the tour origin. The structure is similar to those used for the various tour models, only without the time-of-day sub-model (when the stop occurs is fixed by the tour start/end time). In the stops model, each tour is treated as an independent entity, and once the stop frequency is chosen, each stop is treated independently. Regardless of the tour purpose, the structure of the stops model consists of the following steps:
Systems Analysis Group, WSP USA 2018