- Citilabs recommends including Automated Vehicles (AVs) in regional transportation models
- AVs could increase vehicle availability and trips within households that own an AV, but it could decrease as other modes of transportation become more attractive
- Vary vehicle availability using household/population synthesis
- Vary vehicle trips using trip generation/tour frequency
- AVs have the ability to increase roadway capacity, but automaker risk aversion could decrease it
- Vary roadway capacity using traffic assignment
- AVs used in Transportation Network Companies (TNCs), delivery and freight are likely to increase the number of vehicles on the road at all times of day
- Add trip or tour-based models to represent the movement of TNCs, freight, and delivery vehicles.
Citilabs has built the world’s leading transportation modeling and simulation software, Cube, used in 2,500 cities around the world. Cube predicts the number of vehicles or people that will use a specific transportation facility in the future.
Fully autonomous vehicles (SAE Level 5 AVs) have the potential to disrupt city, region, and state transportation and land use plans. Citilabs recommends for all models that predict transportation beyond 2030 to include AVs. Including AVs in regional models will enable the simulation of multiple scenarios of AV adoption. This blog discusses potential changes caused by AVs and provides tips on how to incorporate possible changes into existing models.
Possible Changes Caused by AVs
Vehicle availability is commonly used to include vehicles that are owned, leased, rented (long term), borrowed from your neighbor, or provided by an employer. It refers to a motor vehicle that is “sitting" in the household driveway. This does not refer to taxis, transportation network company (TNC) vehicles, or other vehicles obtained from a sharing or hourly rental service.
Why availability may increase: Today, most personal automobiles are parked 95% of the time. While not in use by its owner, an AV can perform trips to pick up things, go for maintenance, or be shared for other family members to use.
Why availability may decrease: Personal vehicle ownership has been declining. This decline has been caused by attractive alternative modes of transportation like walking, biking and TNCs. Autonomous vehicles will allow TNCs to use personal vehicles while they are not in use by their owners to move their customers or things for their customers. If sharing a personal vehicle with a TNC does not offset the cost of AV ownership, there could be an even greater decline of personal auto availability.
Different levels of AV usage in metro Atlanta Citilabs setup internal what-if assignments using Citilabs Streetlytics trips and HERE road networks, along with assumptions about the capacity impacts of automated vehicles. This is a series of maps using Citilabs Cube showing congestion associated with different AV adoption rates.
A vehicle trip is a journey between two points made by a single privately-owned vehicle regardless of the number of persons in the vehicle.
Why vehicle trips may increase: In 2009, There were 685 drivers for every 1,000 residents in the US. AV’s will enable most of the other 315 residents to use an automobile by themselves. There will be few limitations for children, the elderly, the disabled, the family groceries, or a vacant car to be on the road.
AVs can decrease the mental cost of driving. AVs can offer door-to-door service eliminating the chore of parking and the terminal time it takes to find a parking spot and to walk to the destination. AVs have the potential to lower or even eliminate these traditional parking costs by finding a cheap parking spot or just continue to circle until the car is needed again. Finally, AVs allow the driver to become the passenger, so they perform other tasks while in the car.
Why trips may decrease: Today 55% of the global population lives in cities, which will rise to 68% by 2050. This trend to live in more dense communities limits the public space available for roads. These shorter trips, can be served by alternative modes of transportation, potentially resulting in a net reduction in personal vehicular trips.
A slide titled The Unbundling is Coming from a presentation called When Mobility Attacks by Horace Dediu on September 6, 2018.
Roadway capacity is the maximum hourly rate at which vehicles can be reasonably expected to traverse a segment of road during a given period of time in normal traffic conditions. This includes all motor vehicles but excludes bicycles and pedestrians.
Why capacity may increase: In 2015, the National Highway and Traffic Safety Administration reported that 94% of motor vehicle crashes were caused by human error. Not only do AVs remove human error from driving, but they also create a stable traffic flow by reducing weaving and speed irregularity. AVs will be capable of operating at higher speeds and closer together with the possibility of platooning. This capability would allow AVs to accelerate or brake simultaneously allowing for a closer headway between vehicles. Platooning creates greater fuel economy by reducing air resistance and congestion and increasing roadway capacity.
If these vehicles can connect to each other, they can connect with the roadway infrastructure and other objects in an Intelligent Transportation System (ITS). This can benefit the flow of traffic by changing traffic signals for high demand by any specific mode of travel.
Why capacity may decrease: Shifting the driving risk from a human to the automaker might force the automaker to restrict AVs to overly cautious behavior. Automakers would restrict AVs to slower speeds and with larger than normal headways to maximize the reduction of risk of being at fault in a collision. This reaction could be more evident when in the proximity of human drivers.
AVs in TNCs, delivery and freight are likely to increase the number of vehicles on the road at all times. Delivery and freight AVs will not need to be limited to 8 hours a day. Faster delivery time will increase the use of just in time manufacturing as well as internet shopping.
Autonomous vehicles will impact regional travel demand models, but the extent of the impact is unknown. Below are tips on how to incorporate these potential changes as variables into existing models. These variables can be changed to test the impacts of different AV adoption rates within any region.
Types of travel demand models
- Trip-based models – Travel demands are commonly estimated using average characteristics for categories of households. Travel between origins and destinations are represented as generally independent trips.
- Tour-Based Models – Travel demands are commonly estimated using characteristics of individual households. Travel between origins and destinations are represented as a sequence of trips which are a part of a tour. Tours are travel events that start at one location and return to that same location.
- Activity-Based Models – Travel demands are commonly estimated using characteristics of individual people, including their household characteristics, and the travel requirements of other household members. Travel between origins and destinations are presented as a sequence of trips which are part of a tour, often considering joint travel with other household members.
Varying vehicle availability using household / population synthesis
Traditionally, travel demand models begin with a step that identifies the socioeconomic characteristics of individual people or households. Trip-based models often stratify the number of households in a zone by vehicle ownership and tour/activity models include the number of vehicles available to each person or household. Autonomous Vehicles could be included here by adding an additional stratification based on vehicle type: human-driven or AV. The regional adoption rate of AVs can be expressed in terms of the percentage of on-road vehicles that are AVs.
While it is possible to handle AV ownership in a trip-based model, it is significantly easier to do so in a tour-based model. Tour-based models include an early step to estimate the number of vehicles available to the household. It would be fairly simple to extend this to estimate the number of AVs as a function of the household’s likeliness to own an AV plus the regional AV adoption rate. Alternatively, a trip-based model would need to subdivide this step by splitting households by vehicle ownership into two stages: households that own 0, 1, 2, 3+ normal vehicles and households that own 0, 1, 2, 3+ AVs.
CUBE TIP: Include an additional stratification based on vehicle type: human-driven or AV
Varying vehicle trips using tour frequency / trip generation
A new trip rate and tour frequency coefficient could be assigned to the new AV vehicle type. The trip generation/tour frequency step would be modified to adjust total trip-making by a percentage for those households that own an AV. This percentage could increase for households with multiple cars or decrease for households with alternative modes of transportation available to them. Some models use an “At-Work” purpose, whose trip rate could be increased to account for empty vehicles moving to a more attractive parking space or returning home after delivering someone to work.
CUBE TIP: Factor trips made by AVs
Varying roadway capacity using traffic assignment
Until AVs have separation from human-driven traffic, they will probably be able to use the same volume/delay functions that exist in nearly every model.
Some traffic assignment routines use a passenger car equivalence (PCE) factor to represent heavy trucks. The theory is that because of its size and limited acceleration/ deceleration characteristics, a heavy truck is equivalent to 2 to 3 passenger cars in terms of its effect on roadway capacity and congestion. The connected nature of AVs and their superior reaction time could merit the use of a PCE of less than 1. A PCE greater than 1 could represent autonomous automaker’s aversion to risk.
CUBE TIP: Run multiple scenarios and use PILOT to adjust the PCE factor marginally each time to assess sensitivity.
AVs could be assigned separately from other vehicles using a standard multiclass assignment technique. The creation of special roadway lanes that are limited to only AV use could be modelled in the same way in which high-occupancy vehicles are modelled today.
CUBE TIP: Code additional links parallel to existing ones to segregate AVs on high capacity roads while reducing the number of lanes for non-AVs.
AV freight and delivery traffic can be shown by creating a separate trip model to show the movement of light/medium commercial vehicles and tractor trailers. Separate tour models can be created to show light/medium delivery vehicles that make multiple stops for Amazon or UPS.
A similar separate trip or tour-based model can be created to represent taxi or TNC movement even when the vehicle is empty traveling between locations.
CUBE TIP: Create separate trip or tour-based models to represent the movement of TNCs, freight, and delivery vehicles.
The characteristics of individual vehicle types is important for dynamic traffic assignment (DTA)/micro-simulation. It is necessary to define the size, acceleration, and deceleration characteristics of passenger cars, buses, and trucks of different sizes. The characteristics for AVs would need to be defined as well.
Destination choice / gravity model
Gravity model shows the trip interchange between zones depending on the relative attraction of each of the zones and the spatial separation between zones. A key factor influencing the choice of destination is the impedance that separates one location from another. This impedance is a function of travel time and cost. Usually, cost is converted into equivalent time by dividing it by the value of time (VOT). Some modelers convert time into equivalent cost by multiplying time by the value of time. As AVs become more common, the value of and sensitivity to travel time may decrease. This decrease could lead to slightly longer trip times because destinations that are more distant would become slightly more attractive.
CUBE TIP: Run several scenarios with a new AV user class and each time reduce the VOT to test sensitivity.
Mode choice models analyze the modes of transportation available to a person or household and predict which mode they select for each trip. The value of time will change the relative relationship of time and cost coefficients, which would influence the choice of mode. Citilabs recommends the addition of AV, TNC, and light individual transportation (like bicycles and scooters) as possible modes available to households.
CUBE TIP: Add AV, TNC, and light individual transportation as possible modes of travel.
For more information
Autonomous vehicles have the potential to disrupt city, region, and state transportation and land use plans. Citilabs recommends for all models that predict transportation beyond 2030 to include AVs. Citilabs’ Cube is an especially malleable software package capable of simulating and predicting the impact of all possible futures of mobility.
Citilabs has recently developed a version of its sample travel demand model and input data, known as Cubetown. This new model demonstrates the likely changes in travel behavior from the adoption of autonomous vehicles.
Citilabs are experts in the future of mobility and we have a solution team ready to train you on our software, aid in your projects, or run entire projects for you. Contact Citilabs today to find out how we can help you.