Difference between revisions of "CityCast"
|Line 1:||Line 1:|
Latest revision as of 04:19, 9 August 2017
|Documentation||Not yet available|
|Data Input||Consumer data (e.g. Acxiom, Epsilon, Experian, Merkle), firm data (e.g. Dun & Bradstreet, InfoUSA), origin-destination data (e.g. AirSage, StreetLight Data), road network (e.g. Google Maps, HERE, OpenStreetMap), GTFS, NHTS|
|Data Output||Mode-split road network usage tables|
CityCast is a web-based transportation modelling application developed by Transport Foundry. CityCast offers results similar to what an MPO might obtain from an ABM or 4-step model, but uses passively-collected data and a simplified process to make model development faster and less expensive.
One of CityCast’s distinguishing features is the data it uses to run its model. ABMs generally rely on detailed household surveys and on-board transit surveys, which can take a lot of time and money to collect. CityCast relies on preexisting datasets that are predominantly gathered passively.
- Consumer data - data from marketing data-services firms that offer various demographic characteristics matched with home locations (e.g. Acxiom, Epsilon, Experian, Merkle)
- Firm data - data that details where employers are, how many people they employee, and what sector they operate in (e.g. Dun & Bradstreet, InfoUSA)
- Origin-destination data - aggregated cell phone location data (e.g. AirSage, StreetLight Data)
- Network data
- Road network data includes average traffic speeds, indicating congestion or other road issues, developed using cell phone and GPS data (e.g. Google Maps, HERE, OpenStreetMap)
- Transit network information, imported from General Transit Feed Specification datasets
- Parking information is not currently imported, but this is supported by MATSim (see notes below on the role of MATSim) using an extension, so this capability could be feasibly added
- National Household Travel Survey (NHTS) - offers some more information on mode choice, travel patterns, and habits.
The various inputs are fused to create synthesized households and employers. The model proceeds to generate synthesized travel diaries for each person in the simulation based on the observed origin and destination data and NHTS data from similar-sized regions. The trips are then assigned to the network using MATSim.
MATSim begins by assigning each trip to the fastest route, but because certain routes become congested, those routes are not necessarily the fastest for every individual once all trips are assigned. Each individual receives a score, and MATSim searches for an opportunity to improve the score for a given trip. MATSim improves scores by changing routes, changing modes, altering activity start and end times, and many other potential mutations of each person’s daily plan. It continues to optimize plans until everyone has their best possible score..
Scores for the optimization are primarily based on travel time and the time people spend doing activities. Other factors such as cost (for fuel, tolls, etc.) or preference to account for multimodal travel can also be included.
CityCast has not been used for modelling by an MPO at the time of this writing as it is still under development. However, test runs were conducted in Asheville, NC. In those tests, runtime was about 4-5 hours. Runtime can vary greatly, depending predominantly on the size and complexity of the network.
Comparison with an Activity-Based Model (ABM)
See full Activity-Based Model page
ABMs are usually custom built for MPOs. This is usually a very long and expensive project. Running the ABM also requires a significant amount of sample data collected from the region, which also takes a long time and is expensive. CityCast’s platform will be transferable between regions, because it is based on large-sample data products available nationally. Although the data would have to be acquired for each region where it is implemented, the process of synthesizing that data into a population and generating and assigning trips could be carried across regions. This may be a particular boon to smaller regions that may not have the resources to build their own ABM. On the other hand, transferability may mean a loss of sensitivity to a region’s particular transportation tendencies when assigning weights in the model, unless those trends can be derived from the existing datasets.
CityCast offers a less complex and therefore less burdensome model. The simpler model allows for faster runtimes, shorter prep times, and less expertise needed to run the model. On the other hand, lower complexity excludes certain data that may be accounted for in a full ABM, which may then produce somewhat different results.