knitr::opts_chunk$set( collapse = TRUE, comment = "#>", eval = identical(tolower(Sys.getenv("NOT_CRAN")), "true"), out.width = "100%" )
In most cases, transport routing models find either the fastest or the lowest-cost routes that connect places in a given transport network. Sometimes, though, we might want a more sophisticated analysis that considers both the time and monetary costs that public transport passengers have to face. The problem here is that simultaneously accounting for both time and monetary costs is a major challenge for routing models because of the trade-offs between the objectives of minimizing trip duration and cost [@conway2019getting].
To address this problem, r5r
has a function called pareto_frontier()
, which calculates the most efficient route possibilities between origin destination pairs considering multiple combinations of travel time and monetary costs. This vignette
uses a reproducible example to demonstrate how to use pareto_frontier()
and
interpret its results.
pareto_frontier
means.Imagine a hypothetical journey from A to B. There are multiple route alternatives between this origin and destination with varying combinations of travel time and cost (figure below).
This figure illustrates the Pareto frontier of alternative routes from A to B. In other words, it shows the most optimal set of route alternatives between A and B. There are certainly other route options, but there is no other option that is both faster and cheaper at the same time.
library(ggplot2) # data.frame df <- structure(list(option = c(1, 2, 3, 4, 5), modes = c("Walk", "Bus","Bus + Bus", "Subway", "Bus + Subway"), time = c(50, 35, 29, 20, 15), cost = c(0, 3, 5, 6, 8)), class = "data.frame", row.names = c(NA, -5L)) # figure ggplot(data=df, aes(x=cost, y=time, label = modes)) + geom_step(linetype = "dashed") + geom_point() + geom_text(color='gray30', hjust = -.2, nudge_x = 0.05, angle = 45) + labs(title='Pareto frontier of alternative routes from A to B', subtitle = 'Hypotetical example') + scale_x_continuous(name="Travel cost (BRL)", breaks=seq(0,12,3)) + scale_y_continuous(name="Travel time (minutes)", breaks=seq(0,60,10)) + coord_cartesian(xlim = c(0,14), ylim = c(0, 60)) + theme_classic()
This kind of abstraction allows us to have a better grasp of the trade-offs between travel time and monetary cost passengers face when using public transport. It also allows us to calculate cumulative-opportunity accessibility metrics with cutoffs for both time and cost (e.g. the number of jobs reachable from a given origin with limits of 40 minutes and $5) (ref paper by Matt and Anson).
Let's see a couple concrete examples showing how r5r
can calculate the Pareto
frontier for multiple origins.
pareto_frontier()
.setup_r5()
First, let's build the network and create the routing inputs. In this example
we'll be using the a sample data set for the city of Porto Alegre (Brazil)
included in r5r
.
# increase Java memory options(java.parameters = "-Xmx2G") # load libraries library(r5r) library(data.table) library(ggplot2) library(dplyr) # build a routable transport network with r5r data_path <- system.file("extdata/poa", package = "r5r") r5r_core <- setup_r5(data_path) # routing inputs mode <- c('walk', 'transit') max_trip_duration <- 90 # minutes # load origin/destination points of interest points <- fread(file.path(data_path, "poa_points_of_interest.csv"))
Now we need to set what are the fare rules of our public transport system. These
rules will be used by R5
to calculate the monetary cost of alternative routes.
In the case of Porto Alegre, the fare rules are as follows:
Each bus ticket costs R$ 4.80.
Riding a second bus adds $
2.40 to the total cost. Subsequent bus rides cost the full ticket price of $ 4.80.
Each train ticket costs $ 4.50. Once a passenger enters a train station, she can take an unlimited amount of train trips as long as she doesn’t leave a station.
The integrated fare between bus and train has a 10% discount, which totals $ 8.37.
We can create list
object with these fare rules with the support of the setup_fare_structure()
function as shown in the code below. A detailed explanation of how to use the fare structure of 5r5
can be found in (this other vignette).
# create basic fare structure fare_structure <- setup_fare_structure(r5r_core, base_fare = 4.8, by = "MODE") # update the cost of bus and train fares fare_structure$fares_per_type[, fare := fcase(type == "BUS", 4.80, type == "RAIL", 4.50)] # update the cost of tranfers fare_structure$fares_per_transfer[, fare := fcase(first_leg == "BUS" & second_leg == "BUS", 7.2, first_leg != second_leg, 8.37)] # update transfer_time_allowance to 60 minutes fare_structure$transfer_time_allowance <- 60 fare_structure$fares_per_type[type == "RAIL", unlimited_transfers := TRUE] fare_structure$fares_per_type[type == "RAIL", allow_same_route_transfer := TRUE]
For convenience, we can save these fare rules as a zip
file and load again for
a future application.
# save fare rules to temp file temp_fares <- tempfile(pattern = "fares_poa", fileext = ".zip") r5r::write_fare_structure(fare_structure, file_path = temp_fares) fare_structure <- r5r::read_fare_structure(file.path(data_path, "fares/fares_poa.zip"))
pareto_frontier()
.In this example, we calculate the Pareto frontier from all origins to all destinations considering multiple cutoffs of monetary costs: - $1, which would only allow for walking trips - $4.5, which would only allow for rail trips - $4.8, which would allow for a single bus trip - $7.20, which would allow for bus + bus - $8.37, which would allow for walking walking + bus + rail
departure_datetime <- as.POSIXct("13-05-2019 14:00:00", format = "%d-%m-%Y %H:%M:%S") prtf <- pareto_frontier(r5r_core, origins = points, destinations = points, mode = c("WALK", "TRANSIT"), departure_datetime = departure_datetime, fare_structure = fare_structure, fare_cutoffs = c(1, 4.5, 4.8, 7.20, 8.37), progress = TRUE ) head(prtf)
For the sake of illustration, let's check the optimum route alternatives from the Farrapos train station to (a) the Praia de Belas shopping mall and (b) the Moinhos hospital. An optimum route alternative means that one cannot make a faster trip without increasing costs, and one cannot make a cheaper trip without increasing travel time.
# select origin and destinations pf2 <- dplyr::filter(prtf, to_id == 'farrapos_station' & from_id %in% c('moinhos_de_vento_hospital', 'praia_de_belas_shopping_center')) # recode modes pf2[, modes := fcase(monetary_cost == 1, 'Walk', monetary_cost == 4.5, 'Train', monetary_cost == 4.8, 'Bus', monetary_cost == 7.2, 'Bus + Bus', monetary_cost == 8.37, 'Bus + Train')] # plot ggplot(data=pf2, aes(x=monetary_cost, y=travel_time, color=from_id, label = modes)) + geom_step(linetype = "dashed") + geom_point() + geom_text(color='gray30', hjust = -.2, nudge_x = 0.05, angle = 45) + labs(title='Pareto frontier of alternative routes from Farrapos station to:', subtitle = 'Praia de Belas shopping mall and Moinhos hospital', color='Destination') + scale_x_continuous(name="Travel cost ($)", breaks=seq(0,12,2)) + scale_y_continuous(name="Travel time (minutes)", breaks=seq(0,120,20)) + coord_cartesian(xlim = c(0,12), ylim = c(0, 120)) + theme_classic() + theme(legend.position=c(.2,0.8))
r5r
objects are still allocated to any amount of memory previously set after they are done with their calculations. In order to remove an existing r5r
object and reallocate the memory it had been using, we use the stop_r5
function followed by a call to Java's garbage collector, as follows:
r5r::stop_r5(r5r_core) rJava::.jgc(R.gc = TRUE)
If you have any suggestions or want to report an error, please visit the package GitHub page.
Any scripts or data that you put into this service are public.
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.