---
title: 'Travel time matrices'
date: "`r Sys.Date()`"
output: rmarkdown::html_vignette
abstract: "This vignette shows how to use the travel_time_matrix() and expanded_travel_time_matrix() functions in r5r."
urlcolor: blue
vignette: >
  %\VignetteIndexEntry{Travel time matrices} 
  %\VignetteEngine{knitr::rmarkdown} 
  \usepackage[utf8]{inputenc}
bibliography: references.json
---

```{r, include = FALSE}
knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  eval = identical(tolower(Sys.getenv("NOT_CRAN")), "true"),
  out.width = "100%"
)

## removes files previously created by 'build_network()'
#data_path <- system.file("extdata/poa", package = "r5r")
#existing_files <- list.files(data_path)
#files_to_keep <- c(
#  "poa_hexgrid.csv",
#  "poa_osm.pbf",
#  "poa_points_of_interest.csv",
#  "poa_eptc.zip",
#  "poa_trensurb.zip",
#  'fares'
#  )
#files_to_remove <- existing_files[! existing_files %in% files_to_keep]
#invisible(file.remove(file.path(data_path, files_to_remove)))
```

# 1. Introduction 

Some of the most common tasks in transport planning and modeling involve require having good quality data with travel time estimates between origins and destinations. `R5` is incredibly fast in generating realistic door-to-door travel time estimates in multimodal transport systems.

The `r5r` package has two functions that allow users to leverage the computing power of `R5`:
- `travel_time_matrix()`
- `expanded_travel_time_matrix()`
- `arrival_travel_time_matrix()`

This vignette shows a reproducible example to explain how these two functions work and the differences between them.


# 2. Build routable transport network with `build_network()`

First, let's build the multimodal transport network we'll be using in this vignette. In this example we'll be using the a sample data set for the city of Porto Alegre (Brazil) included in `r5r`.


```{r, message = FALSE}
# increase Java memory
options(java.parameters = "-Xmx2G")

# load libraries
library(r5r)
library(data.table)
library(ggplot2)

# build a routable transport network with r5r
data_path <- system.file("extdata/poa", package = "r5r")
r5r_network <- build_network(data_path)

# routing inputs
mode <- c('walk', 'transit')
max_trip_duration <- 60 # minutes

# departure time
departure_datetime <- as.POSIXct("13-05-2019 14:00:00", 
                                 format = "%d-%m-%Y %H:%M:%S")

# load origin/destination points
points <- fread(file.path(data_path, "poa_points_of_interest.csv"))

```


# 3. The `travel_time_matrix()` function

The `travel_time_matrix()` function provides a simple and really fast way to calculate the travel time between all possible origin destination pairs at a given departure time using a given transport mode.

The user can also customize many parameters such as:
- `max_trip_duration`: maximum trip duration
- `max_rides`: maximum number of transfer in the public transport system
- `max_walk_time` and `max_bike_time`: maximum walking or cycling time to and from public transport
- `walk_speed` and `bike_speed`: maximum walking or cycling speed
- `max_fare`: maximum monetary cost in public transport. [See this vignette](https://ipeagit.github.io/r5r/articles/fare_structure.html).


```{r, message = FALSE}
# estimate travel time matrix
ttm <- travel_time_matrix(r5r_network,   
                          origins = points,
                          destinations = points,    
                          mode = mode,
                          max_trip_duration = max_trip_duration,
                          departure_datetime = departure_datetime )

head(ttm, n = 10)

```

Now remember that travel time estimates can vary significantly across the day because of variations in public transport service levels. In order to account for this, you might want to calculate multiple travel time matrices departing at different times.

This can be done very efficiently by using the `time_window` and `percentile` parameters in the `travel_time_matrix()` function. When these parameters are set, R<sup>5</sup> will automatically compute multiple travel times estimates considering multiple departures per minute within the `time_window` selected by the user. [More information about this functionality can found in this vignette](https://ipeagit.github.io/r5r/articles/time_window.html).


# 4. The `expanded_travel_time_matrix()` function

Sometimes, we want to know more than simply the total travel time from A to B. This is when the `expanded_travel_time_matrix()` function comes in. By default, the output of this function will also tell which public transport routes were taken between each origin destination pair.

Nonetheless, you may set the parameter `breakdown = TRUE` to gather much more info for each trip. In this case, `expanded_travel_time_matrix()` will tell the number of transfers used to complete each trip and their total access, waiting, in-vehicle and transfer times. Please note that setting `breakdown = TRUE` can make the function slower for large data sets.

*A general call to expanded_travel_time_matrix()*
```{r, message = FALSE}
ettm <- expanded_travel_time_matrix(r5r_network,   
                                    origins = points,
                                    destinations = points,    
                                    mode = mode,
                                    max_trip_duration = max_trip_duration,
                                    departure_datetime = departure_datetime )

head(ettm, n = 10)
```

*Calling expanded_travel_time_matrix() with `breakdown = TRUE`*

```{r, message = FALSE}
ettm2 <- expanded_travel_time_matrix(r5r_network,   
                                    origins = points,
                                    destinations = points,    
                                    mode = mode,
                                    max_trip_duration = max_trip_duration,
                                    departure_datetime = departure_datetime,
                                    breakdown = TRUE)

head(ettm2, n = 10)
```

You will notice in the documentation that the `expanded_travel_time_matrix()` also has a `time_window` parameter. In this case, though, when the user sets a `time_window` value, the `expanded_travel_time_matrix()` will return the fastest route alternative departing each minute within the specified time window. Please note this function can be very memory intensive for large data sets and time windows.

```{r, message = FALSE}
ettm_window <- expanded_travel_time_matrix(r5r_network,   
                                           origins = points,
                                           destinations = points,    
                                           mode = mode,
                                           max_trip_duration = max_trip_duration,
                                           departure_datetime = departure_datetime,
                                           breakdown = TRUE,
                                           time_window = 10)

ettm_window[15:25,]
```



# 4. The `arrival_travel_time_matrix()` function

Both of the functions `travel_time_matrix()` and `expanded_travel_time_matrix()` consider a **departure** time set by the user. In some cases, though, you might need to calculate travel times considering an **arrival time**. For such cases, you can use `arrival_travel_time_matrix()`. In this function, you need to set the latest arrival time desired and a maximum trip duration. The function returns the travel time of the trip with the latest departure time that arrives before the arrival time.

This function is useful when modeling user behavior in situations where arriving by a specific time is important, such as getting to work or school for a set start time (e.g., 9 a.m.). In these scenarios, it is often most convenient for a person to take the latest possible departure that still ensures arrival before their required start time, rather than choosing the trip with the shortest travel time but arriving much earlier and waiting unnecessarily at the destination.

Note that the output of this function includes more information with additional columns, like in the `expanded_travel_time_matrix()` function (you can also use `breakdown = TRUE`).

*A general call to arrival_travel_time_matrix()*
```{r, message = FALSE}

arrival_datetime <- as.POSIXct(
 "13-05-2019 14:00:00",
 format = "%d-%m-%Y %H:%M:%S"
)

arrival_ttm <- arrival_travel_time_matrix(
  r5r_network,
  origins = points,
  destinations = points,
  mode = c("WALK", "TRANSIT"),
  arrival_datetime = arrival_datetime,
  max_trip_duration = 60
)

head(arrival_ttm, n = 10)
```



### Cleaning up after usage

`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:

```{r, message = FALSE}
r5r::stop_r5(r5r_network)
rJava::.jgc(R.gc = TRUE)
```

```{r, eval = TRUE, include = FALSE, message = FALSE}
# clean cache (CRAN policy)
r5r::r5r_cache(delete_file = 'all')

```

If you have any suggestions or want to report an error, please visit [the package GitHub page](https://github.com/ipeaGIT/r5r).

## References