Analyst

Analyst Features

The Analyst adds the following features to OTP:

1. PointSets - A more efficient way to specify a batch of fromPlace / toPlaces for routing.
2. Surfaces - A efficient way to get travel time from one to many places.
3. Isochrones - An alternative implementation of the Isochrone features.

Loading OTP with the Analyst

The analyst is an optional feature in OTP 1.x and can be loaded in otp_setup.

# Get OTP and the Demo Data in the normal way
library(opentripplanner)
path_data <- file.path(tempdir(), "OTP")
dir.create(path_data)
path_otp <- otp_dl_jar(version = "1.5.0") #Must use OTP 1.x
otp_dl_demo(path_data)
log1 <- otp_build_graph(otp = path_otp, dir = path_data)

# Setup OTP with the analyst and pointsets enabled
log2 <- otp_setup(otp = path_otp, dir = path_data, analyst = TRUE, pointsets = TRUE)
otpcon <- otp_connect(timezone = "Europe/London")

You can see the analyst is working in the web UI.

OTP Web Interface in Analyst Mode

Creating a Pointset

You will have noticed that when we started OTP we also enabled the Pointsets feature. Pointsets are .csv files of latitude/longitude coordinates that OTP can route to. PointSets can be created from SF objects using the otp_poinset() function.

We will use the LSOA points from the Advanced Features Vignette to create a PointSet. We will also add a column of data called jobs to analyse. For any numerical values in a PointSet OTP can provide the count and sum of that value based on travel criteria. In this case we might want to know how many jobs can be accessed within a 30 minute drive of a given location.

download.file("https://github.com/ropensci/opentripplanner/releases/download/0.1/centroids.gpkg",
              "centroids.gpkg", mode = "wb")
lsoa <- sf::st_read("centroids.gpkg", stringsAsFactors = FALSE)
lsoa$jobs <- sample(100:500, nrow(lsoa))
otp_pointset(lsoa, "lsoa", path_data)

Here the otp_pointset() function takes the lsoa object and creates a PointSet that can be accessed by OTP using the name “lsoa”.

Creating a Surface

A surface evaluates the travel times from a given point to all locations within 120 minutes in one minute increments. So it quite similar to an isochrone.

The fist step is to make a surface from a given location using otp_make_surface

surfaceid <- otp_make_surface(otpcon, c(-1.17502, 50.64590))

The otp_make_surface function returns a list containing the surface ID and the parameters used to create the surface. You can use this object in other functions that support surfaces.

Once the surface has been made we can evaluate the travel times to our pointset.

ttimes <- otp_surface(otpcon, surfaceid, "lsoa")

The otp_surface function returns a list of two objects. The first object data is a data frame summarising the sum and count of any values in the pointset in one minute increments from the fromPlace specified by the surfaceid object. The second object times is a vector of travel times in seconds to each of the locations in the pointset. NA values are returned if OTP was unable to find a route to the point or if the point is more than 120 minutes from the fromPlace.

If you are not using the sum and count features you can specify get_data = FALSE to get just the travel times.

ttimes <- otp_surface(otpcon, surfaceid, "lsoa", get_data = FALSE)

We can visualise the travel times using the tmap package.

lsoa$time <- ttimes[[1]]$times / 60
library(tmap)
tmap_mode("view")
tm_shape(lsoa) +
  tm_dots(col = "time")

Travel times to LSOA points

Notice that there are a few locations where OTP has failed to find a travel time. It is likely that point is too far from the road network to route, or the mode does not allow access (e.g. driving up a pedestrianised road). You could try modifying your point locations slightly to allow OTP to find a valid route.

Producing a travel time matrix

We can use surfaces as a more efficient way to produce a travel time matrix than using otp_plan. A special function otp_traveltime creates batches of surfaces and pointsets and summarises the results as a travel time matrix.

ttmatrix <- otp_traveltime(otpcon, 
                           path_data, 
                           fromPlace = lsoa,
                           toPlace = lsoa,
                           fromID = lsoa$geo_code,
                           toID = lsoa$geo_code)

The function will return the travel times in seconds between the fromPlaces and toPlaces as a data frame. With columns names from fromID and row names from toID. Any failed routes will return and NA, as will any greater than 120 minutes.

Producing an isochrone from a surface

The surfaces can also be used to produce isochrones from a surface. Only one raster can be produced at a time. If you provide a list of surfaces only the first will be used.

isochone <- otp_surface_isochrone(otpcon, surfaceid)

The function otp_surface_isochrone returns a raster image with the value of the travel time in minutes. You can view the raster using the raster package.

library(terra)
plot(isochone)

Isochone Raster