predictive analytics for transportation in a high...

COLLABORATE. INNOVATE. EDUCATE.

Predictive Analytics for

Transportation Planning and

Operations in a World of Big Data

Chandra R. Bhat

The University of Texas at Austin

Acknowledgments: D-STOP, TxDOT (Jianming Ma+), NCTCOG, Humboldt Award, Dr. Ram Pendyala, Dr. Kostas Goulias, Dr. Abdul Pinjari, all my graduate/undergraduate students


World of high dimensional

heterogeneous data

Providing accurate traffic information is becoming a major challenge.

Cameras, GPS, cell phone tracking, and probe vehicles are used to supplement the information provided by conventional measurement systems.

Methodologies to combine and aggregate high dimensional heterogeneous data are needed


Connected/Automated Vehicles

(CAVs) and big data

The driverless car of the near future is part of a gigantic data-collection engine.

Vehicles have embedded computers, GPS receivers, short-range wireless network interfaces, in-car sensors, cameras, and internet.

Vehicles interact with Roadside wireless sensor networks, passenger’s wireless devices, and other cars.


Data required to keep a self-driving

car safely on the road

Highly detailed maps information: Shape and elevation of roadways, lane lines, intersections, crosswalks, speed limits, and traffic signals.

Position, speed and intentions of other vehicles and pedestrians.

Position, speed and intentions of unexpected obstacles, such as, jaywalking pedestrians, cars lunching out of hidden driveways, a stop sign held up by a crossing guard, and cyclist making gestures.

What can be inferred from CAVs and

smartphones data

• Where people drive,

• when people drive,

• what route people take,

• where people stop,

• what people put in their car,

• why, how and when people take decisions on the fly and change their activity plan route, and

• detailed crashes data (speed, position, and intention at the moment of the accident).


Self-Driving Vehicle (e.g., Google) Connected Vehicle

AI located within the vehicle AI wirelessly connected to an external communications network

“Outward-facing” in that sensors blast outward from the vehicle to collect information without receiving data inward from other sources

“Inward-facing” with the vehicle receiving external environment information through wireless connectivity, and operational commands from an external entity

AI used to make autonomous decisions on what is best for the individual driver

Used in cooperation with other pieces of information to make decisions on what is “best” from a system optimal standpoint

AI not shared with other entities beyond the vehicle

AI shared across multiple vehicles

A more “Capitalistic” set-up A more “Socialistic” set-up

Two Types of Technology


Regular Traffic Conditions

PRESENT DAY


Icy Patch

PRESENT DAY


Incident

PRESENT DAY


Lane blocking, traffic slow down

PRESENT DAY


Congestion buildup, late lane changes

PRESENT DAY


Congestion propagation to frontage, ramp backed up

PRESENT DAY



V2V


Icy Patch

V2V


Incident: Information propagation

V2V


Preemptive lane changing, freeway exit

V2V


Re-optimization of signal timing, upstream detours

INCIDENT AHEAD TAKE

DETOUR

V2I



AUTONOMOUS


Icy Patch

AUTONOMOUS


Avoidance of icy patch, no incident

AUTONOMOUS


Traffic slowdown, late lane changing, congestion

AUTONOMOUS


Icy Patch

AUTONOMOUS + V2X


Avoidance of icy patch, no incident

AUTONOMOUS + V2X


Information propagation, preemptive lane changing, freeway exit

AUTONOMOUS + V2V


Re-optimization of signal timing, upstream detours

INCIDENT AHEAD TAKE

DETOUR

AUTONOMOUS + V2I


Traffic safety is an urgent necessity!

5.7 million crashes in the US in 2013* Enormous cost to society

2,310,000

Injuries in the US in 2013*

232,041 (10%)

Injuries in TX**

*: NHTSA, 2014

**: TxDOT, 2014

Property damage Productivity loss Injury Death

Cost almost $1 trillion!!!

26

32,719

Deaths in the US in 2013*

3.382 (10%)

Deaths in TX**

81% of all annual crashes can potentially

be addressed by vehicle-to-vehicle and

vehicle-to-infrastructure

systems***

***: USDOT, 2015


Multidisciplinary approach to improve Collision

Warning/ Avoidance Systems (CW/CA)

CAR-STOP

TRANSPORTATION

WIRELESS MACHINE

LEARNING

Communication

and radar

technologies

Distributed

decision

making

Transportation aspects

27


Proposed Approach

Prior Info Real-time

info Prediction for CW/CA

IMPROVED

SAFETY

Exploit prior info o Driver´s behavior o Typical traffic condition

Joint radar and communications o Waveform / hardware / antenna design o Security needed to prevent spoofing

Collision warning/avoidance algorithms o Smart combination of information o Distributed decision making

SAFE

28


Benefits of sharing raw sensor data

29

Live streaming of lead vehicle’s front sensor info

Front sensors

Exchanging raw sensor data makes overtaking safer

Rural roadway overtaking scenario

https://www.youtube.com/watch?v=6GNGfse9ZK8

https://www.youtube.com/watch?v=6GNGfse9ZK8


Improving safety for 3 types of traffic

configurations

30

Urban intersections [1]

Rural roads [2]

High density of pedestrians and bicyclists [3]

Sources: [1]Google Maps – City of Austin; [2] atzonline.com; [3] dailytexanonline.com


Real-time Fusion

31

Radar Detection

On-vehicle Sensors

V2V Messages


Exploiting additional sources of

information

32

Radar Detection

On-vehicle Sensors

V2V Messages

Road Topology

Previous Behavior

Extra V2V Messaging


Driver A

Driver B

Driver C

Faster left turn

Slower braking

Personal Motion

Predictor …

Driver history

Personal Motion

Predictor

Predicted better with driver history

Storing and making use of driver history

Months 1-9

33


o Maps provide vital information on expected future motion

o Improve map-based prediction algorithm for CW/CA

OpenStreetMap[1]

Google Maps API[2]

Available road maps

Motion Predictor

[1] www.openstreetmap.org [2] developers.google.com/maps

Work Package 1.3.2

Months 4-14

Acquiring and using location context

34


Data collection: equipment


Performance of joint system: blind intersection

scenario

36


Performance of joint system: position accuracy

37


Performance of joint system: pedestrians and

bicyclists

38

•

•

Position of Vehicle

1

(binary/continuous)

Speed of Vehicle 2

(continuous) Position of Vehicle

2

(binary/continuous)

Speed of

Vehicle 1

(continuous)

Direction

and

angle of

progress

of Veh. 1

Direction

and angle

of

progress

of Veh. 2

Vehicle 1

type/Age

(nominal,

binary)

Vehicle 2

type/age

(nominal,

binary)

Weather

conditions

Convergence

rate index

Vehicle

separability

index

Crash

Occurrence

(yes/no)

Data Science

• Not enough humans to process

• Machine learning, visualization, and advanced computation techniques

• Statistics, social sciences, and domain knowledge

• High-dimensional heterogeneous data


Infrastructure Needs/Planning

Driven By…

Complex activity-travel patterns

Growth in long distance travel demand

Limited availability of land to dedicate to infrastructure

Budget/fiscal constraints

Energy and environmental concerns

Information/ communication technologies (ICT) and mobile platform advances

Autonomous vehicles leverage technology to increase flow without the need to expand capacity


Technology and Infrastructure

Combination Leads To…

Safety enhancement Virtual elimination of driver error – factor in 80% of crashes

Enhanced vehicle control, positioning, spacing, speed,

harmonization

No drowsy, impaired, stressed, or aggressive drivers

Reduced incidents and network disruptions

Offsetting behavior on part of driver


Capacity enhancement Platooning reduces headways and improves flow at transitions

Vehicle positioning (lateral control) allows reduced lane widths and

utilization of shoulders; accurate mapping critical

Optimized route choice

Energy and environmental benefits

Increased fuel efficiency and reduced pollutant emissions

Clean fuel vehicles

Car-sharing


BUT LET’S NOT FORGET

TRAVELER BEHAVIOR

ISSUES!


Impacts on Land-Use Patterns

Live and work farther away Use travel time productively Access more desirable and higher paying job Attend better school/college

Visit destinations farther away

Access more desirable destinations for various activities

Reduced impact of distances and time on activity participation

Influence on developers

Sprawled cities? Impacts on community/regional planning and

urban design


Impacts on Household Vehicle Fleet

Potential to redefine vehicle ownership No longer own personal vehicles; move toward car sharing enterprise where

rental vehicles come to traveler

More efficient vehicle ownership and sharing scheme may reduce the

need for additional infrastructure Reduced demand for parking

Desire to work and be productive in vehicle

More use of personal vehicle for long distance travel

Purchase large multi-purpose vehicle with amenities to work and play in

vehicle


Impacts on Mode Choice

Automated vehicles combine the advantages of public transportation with that of traditional private vehicles

Catching up on news Texting friends Reading novels

Flexibility Comfort Convenience

What will happen to public transportation?

Also automated vehicles may result in lesser walking and bicycling shares

Time less of a consideration So, will Cost be the main policy tool to influence behavior?


Impacts on Mode Choice

Driving personal vehicle more convenient and safe

Traditional transit captive market segments now able to use auto (e.g., elderly, disabled)

Reduced reliance/usage of public transit?

However, autonomous vehicles may present an opportunity for public transit and car sharing Lower cost of operation (driverless) and can cut out low volume routes

More personalized and reliable service - smaller vehicles providing demand-responsive transit

service

No parking needed – kiss-and-ride; no vehicles “sitting” around

20-80% of urban land area can be reclaimed

Chaining may not discourage transit use


Activity Chaining Issues

Drive Alone

Drive Alone Drive Alone

Shopping

Home Work

Very Good Transit Service


Impacts on Long Distance Travel

Less incentive to use public transportation?

Should we even be investing in high capital high-speed rail systems?

Individuals can travel and sleep in driverless cars

Individuals may travel mostly in the night

Speed difference?


Impacts on Commercial Vehicle

Operations

Enhanced efficiency of commercial vehicle operations

Driverless vehicles operating during off-peak and night hours reducing congestion

Reduced need for infrastructure


Mixed Vehicle Operations

Uncertainty in penetration rates of driverless cars

Considerable amount of time of both driverless and traditional car operation

When will we see full adoption of autonomous? Depends on regulatory policies

Need infrastructure planning to support both, with intelligent/dedicated infrastructure for driverless


Concerns about Autonomous Cars

Survey with 1800 individuals in the Puget sound Region

Type of concern Not concerned Somewhat

unconcerned

Neutral/doesn’t

know

Somewhat

concerned Very concerned

Equipment and system

safety 6.9% 4.4% 22.2% 26.9% 39.6%

System and vehicle

security 8.4% 5.0% 26.2% 26.8% 33.7%

Capability to react to

the environment 6.2% 3.2% 18.9% 22.8% 48.9%

Performance in poor

weather or other

unexpected conditions

6.3% 4.3% 21.5% 26.5% 41.4%

Legal liability for drivers

or owners 6.4% 4.2% 24.3% 27.4% 37.7%

DOMAIN KNOWLEDGE IN THE CONTEXT

OF TRAVEL DEMAND MODELING

Exogenous Variable Vector X

Model

• Conceptual/Theoretical/ Methods/Tools and Techniques

• Specification and Definition of Alternatives

Activity-Based Model (ABM) Trip-based Model (TBM)

Outputs

• ABM should…

Capture the central role of activities, time, and space in a continuum

Explicitly recognize constraints and interactions

Represent simultaneity in behavioral choice processes

Account for heterogeneity in behavioral decision hierarchies

Incorporate feedback processes to facilitate integration with land use and network models

• SimAGENT does it all …

SIMAGENT (SIMULATOR OF ACTIVITIES,

GREENHOUSE GAS EMISSIONS, ENERGY,

NETWORKS, AND TRAVEL)

HIGH DIMENSIONAL HETEROGENEOUS

DATA

Why joint modeling of data is important?

• Borrows information on other outcomes

• Able to answer intrinsically multivariate questions, such as the effect of a covariate on a multidimensional outcome

• Obviates the need for multiple tests and facilitates global tests, offering superior testing power and better control of Type 1 error rates

• If some endogenous outcomes are used to explain other endogenous outcomes, and if the outcomes are not modeled jointly, the result can be inconsistent estimation of the effects of one endogenous outcome on another.

• Problem? Mixed data, high-dimensional data

Way-Out

• The new Generalized Heterogeneous Data Model (GHDM).

• Correlation across various dimensions (of the dependent variables) are captured using latent constructs.

• Reduces the size of error covariance elements.

• Accommodates all four types of data (dependent variables).

• Dimension of integration is independent of number of latent constructs.

• Bhat’s Maximum Approximate Composite Marginal Likelihood (MACML) estimation approach is used for estimation of GHDM.

• Integrated modeling = Acknowledging the joint nature of decision-making relevant to inter-related outcomes.

• Jointness may arise because of the impact (on the multiple

choice outcomes) of

common underlying exogenous observed variables

common underlying exogenous unobserved variables

combination of both

• Example: Consider residential location, auto-ownership, and activity time-use.

• Households with low income earnings may…

choose to (or are constrained to) locate in neighborhoods with high population density,

have low household auto ownership levels, and

spend less time in recreational episodes

• If the elements causing jointness are solely due to observed exogenous factors, then modeling is straightforward.


Framework for dwelling unit choice and activity-travel behavior


Activity Travel Behavior and Health Indicators

An Example

• New econometric approach for the estimation of joint mixed models that include:

MDC outcome

nominal discrete outcome

count, binary/ordinal outcomes, as well as continuous outcomes

• Specify latent underlying unobserved factors that impact the many outcomes, and generate the jointness among the outcomes

• Reported subjective attitudinal indicators for the latent variables help provide additional information and stability to the model system

• Application: analysis of residential location choice, household auto-ownership choice, as well as time-use choices

• Residential location choice:

nominal discrete choice with four land-use density categories as defined by housing unit density (housing units per square mile) within census blocks

continuous outcome representing average commute distance

• Household auto ownership is a count outcome

• Individual activity time-use by activity type in non-work activities is an MDC outcome

• To our knowledge, this is the first formulation and application of such an integrated model system in the econometric and statistical literature

Data Source & Dependent Variables

• Puget Sound household travel survey (April-June, 2014)

• Dependent variables

Residential density (hh/sq. mile) : less than 750, 750-1,999, 2,000-2,999, 3,000 or above (nominal variable)

Household average commute distance in miles (continuous variable)

Auto ownership (count variable) Out-of-home discretionary activity: Personal business,

shopping, recreation, dining out, social, and serve passenger. In-home is an outside good. (MDC variable)

• Unit of analysis: Household (For MDC: It is the fraction of time spent participating in different activities by the household)

• Initial sample size: 6,036 households

• Final sample size: 3,637

Indicator variable: Ordinal variables

Attitudinal Question

Response

Very

Unimportant

1 2 3 4

Very

Important

5

How important when choosing current

home:

Having a walkable neighborhood and

being near to local activities 5.5% 7.6% 11.1% 32.3% 43.5%

Being close to public transit 15.4% 12.0% 17.0% 24.8% 30.8%

Being within a 30-minute commute to

work 6.6% 6.5% 10.0% 24.4% 52.5%

Quality of schools (K-12) 31.2% 7.5% 26.7% 14.6% 20.0%

Having space and separation from

others 9.2% 13.7% 21.8% 34.3% 21.0%

Being close to the highway 12.7% 16.0% 21.4% 38.0% 11.9%

Dependent variable: Count variable

Motorized

Vehicle Count

Frequency

0 1 2 3 4 5 >6

Number 304 1,378 1,354 413 135 36 17

% 8.4 37.8 37.2 11.4 3.7 1.0 0.5

Dependent variable: MNP variable

Residential Density

(households per sq. mile) Number of observations (%)

<750 478 (13.2)

750-2,000 866 (23.8)

2,000-3,000 525 (14.4)

>3,000 1,768 (48.6)

Dependent variable: MDC variable

Activity Participation (%) Mean

fraction

Number of households (% of

total number) spent time…

Only in activity

type

In other

activity types

too

In home (IH) 3,637 (100.0) 0.780 533 (14.7) 3,104 (85.3)

Personal Business 1,607 ( 44.2) 0.202 216 (13.4) 1,391 (86.6)

Shopping 1,664 ( 45.8) 0.060 355 (21.3) 1,309 (78.7)

Recreation 1,011 ( 27.8) 0.131 148 (14.6) 863 (85.4)

Dining Out 1,092 ( 30.0) 0.081 203 (18.6) 889 (81.4)

Social 659 ( 18.1) 0.180 82 (12.4) 557 (87.6)

Serve Passenger 751 ( 20.6) 0.047 26 ( 3.5) 725 (96.5)

Latent Variables

• Green lifestyle propensity (GLP)

• Luxury lifestyle propensity (LLP)

Indicators

• Green lifestyle propensity (GLP) Average household commute distance (continuous)

Auto ownership (count)

Having a walkable neighborhood (ordinal with five scale)

Being close to public transit (ordinal with five scale)

Being within a 30-minute commute to work (ordinal with five scale)

• Luxury lifestyle propensity (LLP) Auto ownership (count)

Having space and separation from others (ordinal with five scale)

Quality of schools (ordinal with five scale)

Being close to the highway (ordinal with five scale)

Latent Variable Determinants

Green Lifestyle Propensity:

• Lower income households have a higher GLP relative to higher income.

• Households with a high fraction of young adults (less than the age of 34 years) have a higher GLP relative to those with a low fraction of young adults.

• Higher GLP associated with households with a high fraction of women (relative to a low fraction of women) and a high fraction of well-educated individuals in the household (relative to a low fraction of well-educated individuals).

Luxury Lifestyle Propensity:

• LLP increases with household income

• LLP increases with the number of children in the household

• LLP increases with the age of household members

Correlation between GLP and LLP:

Negative correlation (a green lifestyle is associated with conservative consumption of resources, while a luxury lifestyle correlates with extravagant living)

Effect of Latent Constructs

• Households with a GLP disposition

will prefer to reside in high density neighborhoods close to their workplace,

own few or no vehicles,

and engage more in IH activities and OH social and active recreation activities

• Households with an LLP disposition

will be inclined to locate in low to medium density neighborhoods,

own many vehicles,

and potentially be engaged in more OH shopping and dining out activities

Endogenous Effects

• Residential density of the household’s location impacts the household auto ownership level and household activity time use.

Residing in lower (higher) density neighborhoods leads to a higher (lower) auto ownership level and lower (higher) baseline preferences for OH recreational activities, shopping, and dining out.

Time investment in serve passenger activity increases as one moves from the highest density neighborhoods to progressively lower density neighborhoods.

• Commute distance impacts only time use.

Households with longer commute distances spend more time on shopping, recreation, and dining out.

Examining “True” Effects of

Neo-Urbanist Densification Efforts

Variable ATE from GHDM ATE from IHDM

% Difference Attributable to

“True” Effect Self-Selection

Effect Vehicle ownership

-0.143 (0.011) -0.340 (0.021) 42 58

Participation on

Personal business

-0.037 (0.013) -0.041 (0.013) 90 10

Shopping 0.011 (0.004) 0.019 (0.007) 65 35

Recreation 0.134 (0.021) 0.190 (0.014) 71 29

Dining out 0.094 (0.020) 0.119 (0.021) 79 21

Social -0.056 (0.014) -0.078 (0.017) 72 28

Serve Passenger -0.156 (0.033) -0.162 (0.025) 96 4

Table: Treatment Effects Corresponding to Transplanting a Random Household from a Lowest Density Neighborhood (<750 hh/sq. mile) to Highest Density Neighborhood (>3000 hh/sq. mile)

(standard error in parenthesis)

The Bottom Line

Uncertainty, Uncertainty, Uncertainty

More uncertainty implies more need for planning

But planning must recognize the uncertainty (need a change in current thinking and philosophy)

Thank you

Contact: [email protected]

Five Pillars of ABM Design

• Based on sound behavioral theory/paradigm

• Computationally feasible and tractable

– Model estimation

– Model implementation

• Optimal use of available data (present and future)

• ABM should be both an Activity-Based Model and an Agent-Based Model

• Sensitive to policy issues and planning applications of interest

Behavioral Basis of ABM

• Decision hierarchies and choice processes

– A variety of behavioral decision structures possible

– Virtually all models assume a sequential decision structure similar to traditional four-step models for computational convenience

• Considerable evidence of simultaneity in behavioral choice mechanisms

– Several choices made simultaneously as a lifestyle package

Behavioral Basis of ABM

• Examples of simultaneous choice packages

– Residential location, vehicle ownership, mode to pre-planned activities (e.g., work)

– Activity type, activity duration, and activity timing (scheduling)

• Behavioral heterogeneity

– Differences in choice processes across market segments

– Identify market segments both exogenously and endogenously (latent market segments)

Time-Space Interactions

Home Work

Activity 1 (Fixed)

Activity 2 (Fixed) Ti

me

Urban Space

1

v

Home Activity

A

Activity at Location A

Activity 1

Activity 2

Agent Interactions

I have a client meeting today; so I will take the

car

I have to pick up Jane from School and go shopping later; I need the

car.

My meeting is in the morning. I can pick up

Jane from school today. And we can go shopping together in the evening. OK, that sounds

good. I’ll go ahead and take

light rail today to work. See you

later.

Hey, Mom and Dad, don’t forget; you have to drop

me off at Johnny’s house in the

evening today

Don’t worry Jane; we’ll drop you off on the way to the store and pick you up later. Run along now,

you’ll miss the bus.

Definition of an Activity

• Disaggregate activity purpose definition

– Challenge traditional notion of mandatory and discretionary activities/trips

– Movie, ball game, and child’s tennis lesson or soccer game often have spatial and/or temporal fixity

– Characterize activities and trips by level of spatial and temporal fixity/constraints (besides purpose)

– Can be accomplished using concepts of time-space geography

– Automated method to add attributes describing degrees of freedom according to set of spatial/temporal fixity criteria to activity records in data set

Central Role of Time Use

• Notion of time is central to activity-based modeling

– Explicit modeling of activity durations (daily activity time allocation and individual episode duration)

– Treat time as “continuous” and not as “discrete choice” blocks

• Activity engagement is the focus of attention

– Travel patterns are inferred as an outcome of activity participation and time use decisions

– Continuous treatment of time dimension allows explicit consideration of time constraints on human activities

• Reconcile activity durations with network travel durations (feedback

processes)

In Summary

• ABM should…

– Capture the central role of activities, time, and space in a continuum

– Explicitly recognize constraints and interactions

– Represent simultaneity in behavioral choice processes

– Account for heterogeneity in behavioral decision hierarchies

– Incorporate feedback processes to facilitate integration with land use and network models

• SimAGENT does it all …

• Individuals who are intrinsically dynamic with an active lifestyle may….

search for high density neighborhoods that offer high accessibility to activity locations,

own fewer autos, and

invest substantial time in recreational pursuits

• When an unobserved factor affects the multiple outcomes,

Independently modeling

→ inefficient estimation of covariate effects for each outcome

→ inconsistent estimation of the effects of one endogenous outcome on another

Thank you

Contact: [email protected]

CEMDAP

A COMPREHENSIVE ECONOMETRIC

MICROSIMULATOR OF DAILY

ACTIVITY-TRAVEL PATTERNS

CEMDAP – The Core ABM in SimAGENT

Socio-Economic Data

PopGen

CEMSELTS

CEMDAP

• Simulates activity schedule and travel characteristics for each individual of the region

• Core module of SimAGENT

• 52 sub-models.

• Developed by UT Austin

Features of CEMDAP (continued)

• Changes in the activity-travel pattern of one individual in a household may bring about changes in activity-travel patterns of other household members

• MDCEV approach facilitates modeling activity participation at a

household level with joint activity participation incorporated in

a simple fashion

– MDCEV – Multiple Discrete Continuous Extreme Value

econometric choice modeling method

• Includes a model of household vehicle ownership by type and make/model, and primary driver assignment

INNOVATION:

COMPREHENSIVE REPRESENTATION OF

INTRA-HOUSEHOLD INTERACTIONS

Joint Activities and Household Interactions

MDCEV Model

• Most activity based models accommodate activity type choice as a series of models for each individual in the household

• These approaches do not explicitly recognize that activity participation is a collective decision of household members

• MDCEV approach – simple and relatively inexpensive for modeling activity participation at a household level

• SimAGENT now features MDCEV modeling methodology to capture household-level activity participation

Joint Activities and Interactions

MDCEV Model

• Conventional discrete choice frameworks need to generate mutually exclusive alternatives results in an explosion in the number of alternatives

• MDCEV allows us to tackle the problem by considering activity participation as a household decision

• MDCEV offers substantial computational and behavioral advantages

– Employ one model to generate activities

– Accommodate substitution/complementarity in activity participation and household member dimensions

MDCEV Model

P1 P2 P1 P2

None None None

A1 None None

A2 None None

A1 A2 None None

P1 P2 P1 P2

None None A1

A1 None A1

A2 None A1

A1 A2 None A1

P1 P2 P1 P2

None None A2

A1 None A2

A2 None A2

A1 A2 None A2

P1 P2 P1 P2

None None A1 A2

A1 None A1 A2

A2 None A1 A2

A1 A2 None A1 A2

P1 P2 P1 P2

None A1 None

A1 A1 None

A2 A1 None

A1 A2 A1 None

P1 P2 P1 P2

None A1 A1

A1 A1 A1

A2 A1 A1

A1 A2 A1 A1

P1 P2 P1 P2

None A1 A2

A1 A1 A2

A2 A1 A2

A1 A2 A1 A2

P1 P2 P1 P2

None A1 A1 A2

A1 A1 A1 A2

A2 A1 A1 A2

A1 A2 A1 A1 A2

P1 P2 P1 P2

None A2 None

A1 A2 None

A2 A2 None

A1 A2 A2 None

P1 P2 P1 P2

None A2 A1

A1 A2 A1

A2 A2 A1

A1 A2 A2 A1

P1 P2 P1 P2

None A2 A2

A1 A2 A2

A2 A2 A2

A1 A2 A2 A2

P1 P2 P1 P2

None A2 A1 A2

A1 A2 A1 A2

A2 A2 A1 A2

A1 A2 A2 A1 A2

P1 P2 P1 P2

None A1 A2 None

A1 A1 A2 None

A2 A1 A2 None

A1 A2 A1 A2 None

P1 P2 P1 P2

None A1 A2 A1

A1 A1 A2 A1

A2 A1 A2 A1

A1 A2 A1 A2 A1

P1 P2 P1 P2

None A1 A2 A2

A1 A1 A2 A2

A2 A1 A2 A2

A1 A2 A1 A2 A2

P1 P2 P1 P2

None A1 A2 A1 A2

A1 A1 A2 A1 A2

A2 A1 A2 A1 A2

A1 A2 A1 A2 A1 A2

Each box represents an alternative

MDCEV Model

A1 P1 A1 P2 A1 P1P2

A2 P1 A2 P2 A2 P1P2

Each box

represents an

alternative

None +

Alternatives - Total 7 alternatives versus 64 in traditional case

INNOVATION:

HOUSEHOLD VEHICLE COMPOSITION

AND DRIVER ASSIGNMENT

Vehicle Type Choice Simulation Component

• Vehicle type choice determines vehicle fleet mix; critical to energy and emissions analysis

• SimAGENT incorporates joint vehicle type choice and primary driver allocation model which jointly determines:

– Multiple vehicle holdings

– Body type (Sub-compact, Compact car, Mid-sized car, Large car, Small SUV, Mid-sized SUV, Large SUV, Van, and Pickup)

– Age (Less than 2 years old, 2 to 3 years old, 4 to 5 years old, 6 to 9 years old, 10 to 12 years old, Older than 12 years)

– Make/model and use (miles)

– Primary driver of each vehicle

Vehicle Holdings and Use

Vehicle Type/ Vintage

33 makes/models

21 makes/models

24 makes/models

25 makes/models

7 makes/models

10 makes/models

23 makes/models

19 makes/models

16 makes/models

12 makes/models

13 makes/models

13 makes/models

23 makes/models

15 makes/models

12 makes/models

23 makes/models

12 makes/models

5 makes/models

6 makes/models

15 makes/models

Coupe Old

Sedan Mid-size New

Sedan Mid-size Old

Sedan Compact Old

Sedan Mini/Subcompact New

Sedan Mini/Subcompact Old

Coupe New

Sedan Compact New

Sedan Large Old

Sedan Large New

Minivan Old

Pickup Truck New

SUV New

SUV Old

Hatchback/Station Wagon New

Hatchback/Station Wagon Old

Pickup Truck Old

Van New

Van Old

Minivan New

Non-motorized vehicles

COMPUTATIONAL TECHNIQUES AND

INTEGRATION POTENTIAL

Portable & Flexible Software Architecture

ODBC

Run-Time Data Objects

Household

Person

Zone Data

LOS Data

Pattern

Tour

Stop

Output Files

Simulation

Coordinator

Modeling Modules

… .

.

.

Decision to Work Model

Work Start/End Time

model

Input

Database

Application

Driver

Data Queries

Zone to Zone

Data

Coordinator

Ability to Integrate and Enhance

• Successfully interfaced with

– Multi-period static assignment (the current four-step approach of SCAG)

– TRANSIMS and MATSim (second by second assignment of people and vehicles on networks), and

• Continuous-time evolutionary framework facilitates real-time dynamic integration of ABM and DTA models

• SimAGENT is successfully implemented in the LA region

• Existing SimAGENT code (CEMDAP, PopGen, CEMSELTS) is open source

• Being implemented currently in the New York region; selected based on behavioral realism and ability to accommodate CAVs

• Elements of system being used for long distance travel modeling by CDOT; UT-Austin working with CDOT

• One may develop the reduced form equations as below:

• Parameters to be estimated

• Latent utility differentials of all non-chosen alternatives with respect to the

chosen alternative

→ →

→ and

where and

Reduced Form Model and Estimation

)((

~

ξηcBcxb

ξηBcxb

ξηαscxbyU

S )

S

SS

)(,(MVN)(

ΣIDENΞ )

QGEQ

ccBcxb~yU

] , , ),Vech(),Vech(),Vech(),Vech([ δΣλ θφcbα

qggqgg qgmqgimqgi UUu

ggmqgImqgmqgqg miuuu

qggqgqg;,...,, 21u

qGqqq uuuu ,...,, 21

,)(

qqq uyyu

)(,...,)(,)( 21

Qyuyuyuyu Ω~,

~MVN

)~

(Β~yu

GEQ

)M BcxbB

(~

MΣIDENΞM Ω )(~

Qcc

Estimation

• Further, partition the vector and matrix which correspond to the mean

and variance of continuous, ordinal and count, and differenced utility elements

as follows:

B~

Ω~

) variablescontinuous of(mean 1)vector( ~~

QHBRB yy

) variablesnominal andcount ordinal, of(mean 1)vector~

( ~~

~~ EQBRB uu

) variablescontinuous theof e(covarianc )matrix( ~~

QHQH yyy RΩRΩ

) variablesnominal andcount ordinal, theof e(covarianc)matrix~~

( ~~

~~~ EQEQ uuu RΩRΩ

) variablesremaining and variablescontinuousbetween e(covarianc)matrix~

( ~~

~~ EQQHyy uu RΩRΩ

y

u

BB

B

y yu

yu u

Ω ΩΩ

Ω Ω

• : MNV density function of dimension QH with mean of and covariance

of

• : integral to evaluate the conditional likelihood of all ordinal, count, and

nominal variable outcomes for all Q individuals

• CML:

• Develop the likelihood function by decomposing the joint distribution of

into a product of marginal and conditional distributions

• given is MNV with mean and variance

Likelihood Function

)~,( uyyu

u~ y )~

(~~~ 1

~~~ yuuu ByBB

yy ΩΩ

uuuu ΩΩΩ-ΩΩ ~1

~~~~~~~

yyy

uplowQHfL ψuψB|y yy

~ Pr)

~,

~()( Ωλ

drffEQ

D

QH

r

),|()~

,~

( ~~~ uuyy BrB|y ΩΩ

)~

,~

( yyB|y ΩQHfyB

~

yΩ~

uplow ψuψ

~ Pr

uplowH

Q

q

Q

qq

CML fL ,qqqq,qqy,qqy,qqqq ψuψB|y

~ Pr)

~,

~()( *2

1

1 1

Ωλ

Empirical Application

Framework

predictive analytics for transportation in a high...

Documents