Mid Term Parameters
The mid-term (day-to-day) simulator handles transportation demand for passengers and goods; it simulates agents’ behavior, which includes their activity and travel patterns. The mid-term represents moving vehicles in aggregate, and routes are generated by behavior-based demand models. The inputs to the model system are based on the outcome of SimMobility Long-Term (SimMobilityLT).
The SimMobility mid-term architecture comprises three interconnecting components, namely the Pre-day, Within-day and supply simulators. Models within these simulators are developed utilizing the various datasets from Singapore. Pre-day simulator is responsible for generating the day activity schedule, modelling and simulating individual daily activity travel pattern using 22 models. Within-Day simulator simulates departure times and the route choice behavior incorporating en-route decisions (re-scheduling). Supply simulator takes care of network attributes and supply system in relation to both public and private transport. SimMobilityMT is an open source framework and aims to simulate millions of agents for a large network, which requires very high computing capabilities. Developed in C++, the performance improvements rely on parallel processing and distributed computing through boost threads and boost mpi, respectively. A lightweight, embeddable scripting language LUA is used for model specification.
Day Pattern level
The pattern choice is a function of many types of household characteristics and personal demographic variables. It predict the occurrence of tours and intermediate stops. The occurrence is expressed as a binary variable (0 or 1+). Noticed that this model is the very first model of benchmark models, the only available information is personal and household characteristics.More details can be found in the the Technical Report for Day Pattern Model
Binary model to predict if an individual travels or not (dpb.lua)
A binary logit model predicts if an individual will travel or not. The utility is expressed as a function of alternative-specific variable (cons_travel), characteristics of the individual (employment type,age, adult gender/children, household composition, personal income, work at home_travel, two plus_car, one_car,motorcycle availability, logsums from the binary model for tour). More details can be found in the lua file (dpb.lua).
Model to predict the number of tours (dpt.lua)
A Multinomial Logit model predicts the number of tours. A tour can be simple (work, education, shopping and others), or complex (work and shopping; work and others; education and shopping; education and others; shopping and others; work, education and shopping; education, shopping and others).
The utility expression is expressed as a function of alternative-specific constants, combination of constants, person type based on employment for alternatives, age groups of adults, adult gender per children, household composition, personal income, work at home, zero car ownership, single car ownership, two or more cars ownership, availability of a motorcycle, constants for single tour, two tours and three tours, and logsums. Details about the specification of the model, utility equation and coefficients can be found in dpt.lua.
Model to predict the exact number of stops (dps.lua)
This day pattern model that predicts the exact number of stops was also estimated using Multinomial Logit model. The choice of an individual making a tour with no stop, single stop (work, education, shopping and others), multiple stops (work and shopping; work and others; education and shopping; education and others; shopping and others).
Alternative-specific stop constants, person type, adult age group, adult gender/children, household composition, personal income, work at home, zero car, one car, two or more cars, availability of motorcycle availability, logsums and combination constants are part of the utility equation for the different alternatives. More details could be found in dps.lua.
Tour Level Models
The exact number of tour models will predict the exact number of tour for each of the tour purposes that has value 1 as predicted by day pattern model. There are four models in this section (work tour, education tour, shopping tour and other tour).Please refer to the Technical Report for Exact Number of Tour Model
Model for exact number of tours for education (nte.lua)
The logit model used predicts the exact number of tours. The alternatives considered are 1 and 2. We consider alternative-specific coefficients for employment type, age, adults by gender per child, household composition, personal income, work at home, availability of car, availability of motorcycles and constant for education. Please refer to nte.lua.
Model for exact number of tours for other activites (nto.lua)
A multinomial logit model that predicts the exact number of tours for 'other' activities. The alternatives considered are 1, 2 and 3.
Person type, age group of adults, number of adult based on gender per child, household composition, personal income, work at home, availability of car, availability of motorcycles, logsums and alternative-specific constants are included in the specification. More details about the specification can be obtained from nto.lua.
Model for exact number of tours for shopping (nts.lua)
This binary logit model predicts if an individual will pursue one or two tours for shopping in a given day.
Availability of car, availability of motorcycle, logsums and alternative-specific constant are the parameters used in the utility equation. Please refer to nts.lua for more details.
Model for exact number of tours for work (ntw.lua)
This binary logit model predicts if an individual will pursue one or two tours for shopping in a given day.
Number of adults by gender per child, work at home, availability of car, motorcycle, logsum and alternative-specific constants are included in the utility equation. Please refer to ntw.lua for more details.
Usual/Unusual work place (tuw.lua)
This model will predicate whether an agent with an usual work location will go to usual work location or not for a particular work tour. It should be noted this choice is only for those that has a fixed work location. For those with out fixed work location, the choice is not necessary. The result of this model will determine if it is necessary to predicate work tour location. Please refer to the Technical Report for Binary Choice Between Usual and Unusual Work Location for more details.
Mode choice for education tour (usual location) (tme.lua)
This model will predict the mode choice for education tour. Education tour can only be conducted by full time student (we have such constrains in day pattern model) and the destination has been given. It is only necessary to predict the mode. Please refer to the Technical Report for Mode Choice for Education Tour for more details.
Mode choice for work tour to usual location (tmw.lua)
This model will predict the mode choice for work tour to usual location. For each tour, we need to determine the mode and destination of it. Some agents has fixed or usual work location. Model 2.1 whether these agents will go to the usual location or not. If they go to usual location, then only the mode need to be predicted. That’s when this model comes into action. Please refer to the Technical Report for Mode Choice for Work Tour to Usual Location for more details.
Mode and destination choice for work tour to unusual location (tmdw.lua)
This model will predict both mode and destination for work tours to unusual location. Basically Mode/Destination Choice model for work tour to unusual location, shopping tour, and other tour have the same specification and variables.Please refer to the Technical Report for Mode and Destination Choice for Work Tour to Unusual Location for more details.
Mode and destination choice for shopping tour (tmds.lua)
Model for predicting the mode and destination choice for other tours (tmdo.lua)
Multinomial logit model predict the combined choice of mode and destination choice for other tours. The choice set includes n (modes) * m (destinations) = n * m alternatives.
Variables included in the specification can be divided in two categories, the first including the personal (gender) and household demographic variables (zero cars, one of more cars, two or more cars, three or more cars) and the second including the destination and the mode to destination (travel cost, travel time, central business district, distance, alternative-specific constants). Purpose-specific size (log_size) variables derived from land-use characteristics are included in the model, as attractiveness of the zone. Please refer to tmdo.lua.
Work-based sub-tours
Work-based sub-tour generation (tws.lua)
Model for mode and destination choice for work-based tour (stmd.lua)
Multinomial logit model is used to predict the combined choice of mode and destination for a work-based tour. The choices available will depend on the number of modes and destinations available. For example, for 'n' modes and 'm' destinations, the number of alternatives will be 'n * m'.
Alternative-specific coefficients for cost, travel time, log term for the size, mode used for the work tour and alternative-specific constant are used in the utility equation. Please refer to stmd.lua.
Model for tour time-of-day for other tours (sttd.lua)
Tour time-of-day model for other tours single model, that predicts the choice of the 30 minute time window for pursuing the tour. 48 30-minute time blocks are considered. Multinomial logit model is used for predicting it.
Utility equations are expressed as functions of trigonometric functions and constants. Please refer to sttd.lua
Tour time of day
Tour time of day for education tour (ttde.lua)
Tour time of day for other tour (ttdo.lua)
Tour time of day for work tour (ttdw.lua)
Time of day model at tour level jointly predicts the arrival time and departure time for primary activity in a tour. To model time of day choice in the context of discrete choice model, time is divided into blocks. Please refer to the Technical Report for Tour Time Of Day Model for Work Tour for more details.
Intermediate Stop Level
Model for intermediate stop generation (isg.lua)
A multinomial logit model that predicts if an individual will pursue an intermediate stop for work, education, shopping, others or no intermediate stop at all.
The specification involves the inclusion of alternative-specific constants, constants for first, second and three plus stops in the inbound or outbound leg of the tour, constants for quitting in the first, second or two plus tours. Indicator variables for being female, driver, passenger, travel in a public transport in work tour, education tour, shopping tour, time window of first stop, second stop or 3 or more stops in the first or second leg of work, educational, shopping or other tours are also part of the utility equation. Distance to work, education, shopping and others are included in the expression. Variables indicating stops for the different tour types in the six discrete time intervals (7:00 A.M. to 9:30 A.M., 9:30 A.M. to 12:00 noon, 3:00 P.M. to 5:30 P.M., 5:30 P.M. to 7:30 P.M., 7:30 P.M. to 10:00 P.M. and 10:00 P.M. to 7:00 A.M.) are also part of the utility equation. Kindly refer to isg.lua, and the Technical Report for Intermediate Stop Generation Model
Model to predict the Mode/destination choice for work tour to unusual location (imd.lua)
The multinomial logit model is used to predict the combined choice of mode and destination for a given work tour to an unusual location. The choices available will depend on the number of modes and destinations available. For example, for 'n' modes and 'm' destinations, the number of alternatives will be 'n * m'.
Alternative-specific coefficients specific to modes are used in addition to the alternative-specific constants in this specification. The variables included in the model include travel cost, travel time, dummy variables for work and shopping, CBD Dummy, distance, zero cars, one or more cars, two or more cars, three or more cars, zero motorcycle, one or more motorcycles, two or more motorcycles, three or more motorcycles, gender. Please refer to imd.lua, and the Technical Report for Mode/Destination Choice for Intermediate Stop Model
Model for intermediate stop time of day (itd.lua)
Time-of-day model at the stop level is a single model, applied to stops in both half tours, that predicts the arrival or departure time of stops. This model predicts one time point, which could be the departure time or arrival time depending on whether the stop is in the inbound or outbound half-tour. For stops in the first half-tour, departure time is known (departure time equals arrival time of the next stop or tour primary activity, minus travel time), making it necessary to only predict arrival time. Similar logic could be applied to predict only departure time for second half-tour. 48 30-minute time blocks are considered. For stops in the first half-tour, available alternatives are bound by the time of arrival at home from the previous tour and departure time of current stop. For stops in the second half-tour, available alternatives are bounded by arrival time of the current stop and the end of day. Multinomial logit model is used for predicting it.
Utility equations are expressed as functions of trigonometric functions, constants and income. Please refer to itd.lua, and the Technical Report for Intermediate Stop Tour Time Of Day Model
The within-day module of the mid-term simulator is designed to simulate travelers’ various decision making processes before and during their trips. It has two modes of operation.
- The pathset generation mode, in which, for every Origin-Destination pair involved in the travelers' trips, a set of available paths connecting the origin to the destination is computed.
- The within day normal simulation, where the travelers choose their paths among the available ones and the actual movement trajectories are simulated.
The within-day module is an event driven system, where all the behavior models are triggered by assorted events. From the traveler’s perspective, the simulation flow is driven by a cycle called within-day simulation cycle, which is divided into two distinct states: performing activity and undertaking a trip. When a traveler is performing an activity, incoming events are handled by corresponding pre-trip behavior models; while when a traveler is undertaking a trip, events are handled by en-route behavior models.
Speed-density model
The speed-density function produce the speed value used for individual vehicles on the same segment. This function uses the current density on the road to calculate the speed. The user can change the parameters used in the speed-density formula in the configuration file simrun_MidTerm.xml in the following section:
<speed_density_params>
<param category="1" alpha="3.0" beta="1.5"/>
<param category="2" alpha="3.0" beta="1.5"/>
<param category="3" alpha="3.0" beta="1.5"/>
<param category="4" alpha="3.0" beta="1.5"/>
<param category="5" alpha="3.0" beta="1.5"/>
<param category="6" alpha="3.0" beta="1.5"/>
<param category="7" alpha="3.0" beta="1.5"/>
</speed_density_params>Each link has a property that indicates the category of the speed-density relationship.
Dwell-time model for public bus
The stopping time of the bus, to consent passengers to board and alight, is computed via the dwell-time model. Currently, in the mid-term simulation, the dwell-time is equal to uniform_distribution(minimal_fixed_time, maximal_fixed_time) + max(number of boarding people, number of alighting people) * uniform_distribution(minimal individual time, maximal individual time). The minimal_fixed_time, maximal_fixed_time, minimal individual time, maximal individual time of the formula can be configured in the file simrun_MidTerm.xml as follows:
<dwell_time_parameters>
<parameters value="0.0, 0.0, 2.0, 3.0"/>
</dwell_time_parameters>Capacity setting for public bus
The public buses have finite capacity. Therefore, when passengers want to board onto the arriving bus, the occupation must be checked first. If the arriving bus is full, the waiting person can not board. The capacity setting can be found in the configuration file simrun_MidTerm.xml in the following element:
<bus_default_capacity value="100"/>Pedestrian model
In our model, pedestrians walk at a fixed speed from the original node to the destination node. The walking speed can be configured in the file simrun_MidTerm.xml, in the following element:
<pedestrian_walk_speed value="5.0"/>Public route choice model (ptrc.lua)
The choice set for this model includes the set of all feasible paths for the public transportation system and varies from individual to individual. A multinomial logit model is used.
In vehicle travel times, indicator for walking leg, waiting time, number of transfers, cost for travel and path size logit are part of the utility equation. Please refer to ptrc.lua for more details.
Private route choice (pvtrc.lua)
The choice set for a private vehicle includes the set of all feasible paths for private individual cars. A multinomial logit model is used in the lua functions of the file pvtrc.lua.