Repository: toddwschneider/shiny-salesman
Branch: master
Commit: 229b5bcc41a1
Files: 11
Total size: 15.8 KB

Directory structure:
gitextract_crelo5c4/

├── .gitignore
├── LICENSE
├── README.md
├── data/
│   ├── cities.rds
│   ├── distance_matrix.rds
│   ├── great_circles.rds
│   └── usa_cities.rds
├── helpers.R
├── server.R
├── ui.R
└── www/
    └── custom_styles.css

================================================
FILE CONTENTS
================================================

================================================
FILE: .gitignore
================================================
.Rapp.history
shinyapps

================================================
FILE: LICENSE
================================================
MIT License

Copyright (c) 2017 

Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.


================================================
FILE: README.md
================================================
A Shiny app to solve the traveling salesman problem with simulated annealing. Check out the full post here: https://toddwschneider.com/posts/traveling-salesman-with-simulated-annealing-r-and-shiny/

To run on your local machine, paste the following into your R console:

```R
install.packages(c("shiny", "maps", "geosphere"), repos="http://cran.rstudio.com/")
library(shiny)
runGitHub("shiny-salesman", "toddwschneider")
```

![](http://images.rapgenius.com/0e1ca854cbc30f33abc46108f2ba38f2.640x640x42.gif)


================================================
FILE: helpers.R
================================================
miles_per_meter = 100 / 2.54 / 12 / 5280

if (!exists("all_cities")) all_cities = readRDS("data/cities.rds")
if (!exists("usa_cities")) usa_cities = readRDS("data/usa_cities.rds")

generate_random_cities = function(n = 10, min_dist = 250, usa_only=FALSE) {
  if (usa_only) {
    candidates = usa_cities
  } else {
    candidates = all_cities
  }
  
  cities = candidates[sample(nrow(candidates), 1),]
  candidates = subset(candidates, !(full.name %in% cities$full.name))
  i = 0

  while (nrow(cities) < n & i < nrow(all_cities)) {
    candidate = candidates[sample(nrow(candidates), 1),]
    candidate_dist_matrix = distm(rbind(cities, candidate)[, c("long", "lat")]) * miles_per_meter

    if (min(candidate_dist_matrix[candidate_dist_matrix > 0]) > min_dist) {
      cities = rbind(cities, candidate)
      candidates = subset(candidates, !(candidates$full.name %in% cities$full.name))
    }

    i = i + 1
  }
  
  cities = cities[order(cities$full.name),]
  cities$n = 1:nrow(cities)
  
  return(cities)
}

plot_base_map = function(map_name="world") {
  margins = c(3.5, 0, 3.5, 0)
  if (map_name == "world") {
    map("world", col="#f3f3f3", fill=TRUE, lwd=0.2, mar=margins)
  } else if (map_name == "usa") {
    map("usa", col="#f3f3f3", border=FALSE, fill=TRUE, mar=margins) #, projection="albers", parameters=c(29.5, 45.5))
    map("state", add=TRUE, col="#999999", fill=FALSE) #, projection="albers", parameters=c(29.5, 45.5))
  }
}

plot_city_map = function(cities, map_name="world", label_cities=TRUE) {
  plot_base_map(map_name)
  # TODO: maptools pointLabel() for better label placement
  map.cities(cities, pch=19, cex=1.1, label=label_cities)
}

plot_tour = function(cities, tour, great_circles, map_name="world", label_cities=TRUE) {
  plot_city_map(cities, map_name, label_cities=label_cities)
  
  if (length(tour) > 1) {
    closed_tour = c(tour, tour[1])
    keys = apply(embed(closed_tour, 2), 1, function(row) paste(sort(row), collapse="_"))
    invisible(sapply(great_circles[keys], lines, lwd=0.8))
  }
}

calculate_great_circles = function(cities) {
  great_circles = list()
  if (nrow(cities) == 0) return(great_circles)
  
  pairs = combn(cities$n, 2)
  
  for(i in 1:ncol(pairs)) {
    key = paste(sort(pairs[,i]), collapse="_")
    pair = subset(cities, n %in% pairs[,i])
    pts = gcIntermediate(c(pair$long[1], pair$lat[1]), c(pair$long[2], pair$lat[2]), n=25, addStartEnd=TRUE, breakAtDateLine=TRUE, sp=TRUE)

    great_circles[[key]] = pts
  }
  
  return(great_circles)
}

calculate_tour_distance = function(tour, distance_matrix) {
  sum(distance_matrix[embed(c(tour, tour[1]), 2)])
}

current_temperature = function(iter, s_curve_amplitude, s_curve_center, s_curve_width) {
  s_curve_amplitude * s_curve(iter, s_curve_center, s_curve_width)
}

s_curve = function(x, center, width) {
  1 / (1 + exp((x - center) / width))
}

run_intermediate_annealing_process = function(cities, distance_matrix, tour, tour_distance, best_tour, best_distance,
                                              starting_iteration, number_of_iterations,
                                              s_curve_amplitude, s_curve_center, s_curve_width) {
  n_cities = nrow(cities)
  
  for(i in 1:number_of_iterations) {
    iter = starting_iteration + i
    temp = current_temperature(iter, s_curve_amplitude, s_curve_center, s_curve_width)
    
    candidate_tour = tour
    swap = sample(n_cities, 2)
    candidate_tour[swap[1]:swap[2]] = rev(candidate_tour[swap[1]:swap[2]])
    candidate_dist = calculate_tour_distance(candidate_tour, distance_matrix)

    if (temp > 0) {
      ratio = exp((tour_distance - candidate_dist) / temp)
    } else {
      ratio = as.numeric(candidate_dist < tour_distance)
    }
    
    if (runif(1) < ratio) {
      tour = candidate_tour
      tour_distance = candidate_dist
      
      if (tour_distance < best_distance) {
        best_tour = tour
        best_distance = tour_distance
      }
    }
  }
  
  return(list(tour=tour, tour_distance=tour_distance, best_tour=best_tour, best_distance=best_distance))
}

ensure_between = function(num, min_allowed, max_allowed) {
  max(min(num, max_allowed), min_allowed)
}

seed_cities = c(
  "Buenos Aires, Argentina",
  "Sydney, Australia",
  "Rio de Janeiro, Brazil",
  "Montreal, Canada",
  "Beijing, China",
  "Moroni, Comoros",
  "Cairo, Egypt",
  "Paris, France",
  "Athens, Greece",
  "Budapest, Hungary",
  "Reykjavik, Iceland",
  "Delhi, India",
  "Baghdad, Iraq",
  "Rome, Italy",
  "Tokyo, Japan",
  "Bamako, Mali",
  "Mexico City, Mexico",
  "Kathmandu, Nepal",
  "Oslo, Norway",
  "Port Moresby, Papua New Guinea",
  "Lima, Peru",
  "Kigali, Rwanda",
  "San Marino, San Marino",
  "Singapore, Singapore",
  "Moscow, Russia",
  "Colombo, Sri Lanka",
  "Bangkok, Thailand",
  "Istanbul, Turkey",
  "London, UK",
  "New York, USA"
)


================================================
FILE: server.R
================================================
library(shiny)
library(maps)
library(geosphere)
source("helpers.R")

shinyServer(function(input, output, session) {
  vals = reactiveValues()
  
  map_name = reactive({
    tolower(input$map_name)
  })
  
  set_random_cities = reactive({
    input$set_random_cities + input$set_random_cities_2
  })
  
  city_choices = reactive({
    if (map_name() == "world") {
      return(all_cities)
    } else if (map_name() == "usa") {
      return(usa_cities)
    }
  })
  
  update_allowed_cities = observe({
    if (isolate(input$go_button) == 0 & isolate(set_random_cities()) == 0 & map_name() == "world") return()
    
    updateSelectizeInput(session, "cities", choices=city_choices()$full.name)
  }, priority=500)
  
  one_time_initialization = observe({
    isolate({
      cty = subset(city_choices(), full.name %in% seed_cities)
      cty$n = 1:nrow(cty)
      updateSelectizeInput(session, "cities", selected=cty$full.name)

      vals$cities = cty
      vals$distance_matrix = readRDS("data/distance_matrix.rds")
      vals$great_circles = readRDS("data/great_circles.rds")
    })
  }, priority=1000)
  
  set_cities_randomly = observe({
    if (set_random_cities() == 0 & map_name() == "world") return()
    run_annealing_process$suspend()
    
    isolate({
      if (map_name() == "world") {
        cty = generate_random_cities(n=20, min_dist=500)
      } else if (map_name() == "usa") {
        cty = generate_random_cities(n=20, min_dist=50, usa_only=TRUE)
      }
      
      cty$n = 1:nrow(cty)
      updateSelectizeInput(session, "cities", selected=cty$full.name)
      
      vals$cities = cty
    })
  }, priority=100)
  
  set_cities_from_selected = observe({
    if (input$go_button == 0) return()
    run_annealing_process$suspend()
    
    isolate({
      cty = subset(city_choices(), full.name %in% input$cities)
      if (nrow(cty) == 0 | identical(sort(cty$full.name), sort(vals$cities$full.name))) return()
      cty$n = 1:nrow(cty)
      vals$cities = cty
    })
  }, priority=50)
  
  set_dist_matrix_and_great_circles = observe({
    if (input$go_button == 0 & set_random_cities() == 0 & map_name() == "world") return()
    
    isolate({
      if (nrow(vals$cities) < 2) return()
      if (identical(sort(vals$cities$name), sort(colnames(vals$distance_matrix)))) return()
      
      dist_mat = distm(vals$cities[,c("long", "lat")]) * miles_per_meter
      dimnames(dist_mat) = list(vals$cities$name, vals$cities$name)
      
      vals$distance_matrix = dist_mat
      vals$great_circles = calculate_great_circles(vals$cities)
    })
  }, priority=40)
  
  setup_to_run_annealing_process = observe({
    input$go_button
    set_random_cities()
    map_name()
    
    isolate({
      vals$tour = sample(nrow(vals$cities))
      vals$tour_distance = calculate_tour_distance(vals$tour, vals$distance_matrix)
      vals$best_tour = c()
      vals$best_distance = Inf

      vals$s_curve_amplitude = ensure_between(input$s_curve_amplitude, 0, 1000000)
      vals$s_curve_center = ensure_between(input$s_curve_center, -1000000, 1000000)
      vals$s_curve_width = ensure_between(input$s_curve_width, 1, 1000000)
      vals$total_iterations = ensure_between(input$total_iterations, 1, 1000000)
      vals$plot_every_iterations = ensure_between(input$plot_every_iterations, 1, 1000000)
      
      vals$number_of_loops = ceiling(vals$total_iterations / vals$plot_every_iterations)
      vals$distances = rep(NA, vals$number_of_loops)
      
      vals$iter = 0
    })
    
    run_annealing_process$resume()
  }, priority=20)
  
  run_annealing_process = observe({
    qry = parseQueryString(session$clientData$url_search)
    if (input$go_button == 0 & is.null(qry$auto)) return()
    
    if (nrow(isolate(vals$cities)) < 2) return()
    
    isolate({
      intermediate_results = run_intermediate_annealing_process(
                               cities = vals$cities,
                               distance_matrix = vals$distance_matrix,
                               tour = vals$tour,
                               tour_distance = vals$tour_distance,
                               best_tour = vals$best_tour,
                               best_distance = vals$best_distance,
                               starting_iteration = vals$iter,
                               number_of_iterations = vals$plot_every_iterations,
                               s_curve_amplitude = vals$s_curve_amplitude,
                               s_curve_center = vals$s_curve_center,
                               s_curve_width = vals$s_curve_width
                             )
      
      vals$tour = intermediate_results$tour
      vals$tour_distance = intermediate_results$tour_distance
      vals$best_tour = intermediate_results$best_tour
      vals$best_distance = intermediate_results$best_distance

      vals$iter = vals$iter + vals$plot_every_iterations
      
      vals$distances[ceiling(vals$iter / vals$plot_every_iterations)] = intermediate_results$tour_distance
    })
    
    if (isolate(vals$iter) < isolate(vals$total_iterations)) {
      invalidateLater(0, session)
    } else {
      isolate({
        vals$tour = vals$best_tour
        vals$tour_distance = vals$best_distance
      })
    }
  }, priority=10)
  
  output$map = renderPlot({
    plot_tour(vals$cities, vals$tour, vals$great_circles, map_name=tolower(input$map_name), label_cities=input$label_cities)
    
    if (length(vals$tour) > 1) {
      pretty_dist = prettyNum(vals$tour_distance, big.mark=",", digits=0, scientific=FALSE)
      pretty_iter = prettyNum(vals$iter, big.mark=",", digits=0, scientific=FALSE)
      pretty_temp = prettyNum(current_temperature(vals$iter, vals$s_curve_amplitude, vals$s_curve_center, vals$s_curve_width),
                              big.mark=",", digits=0, scientific=FALSE)
      
      plot_title = paste0("Distance: ", pretty_dist, " miles\n",
                          "Iterations: ", pretty_iter, "\n",
                          "Temperature: ", pretty_temp)
                          
      title(plot_title)
    }
  }, height=550)
  
  output$annealing_schedule = renderPlot({
    xvals = seq(from=0, to=vals$total_iterations, length.out=100)
    yvals = current_temperature(xvals, vals$s_curve_amplitude, vals$s_curve_center, vals$s_curve_width)
    plot(xvals, yvals, type='l', xlab="iterations", ylab="temperature", main="Annealing Schedule")
    points(vals$iter, current_temperature(vals$iter, vals$s_curve_amplitude, vals$s_curve_center, vals$s_curve_width), pch=19, col='red')
  }, height=260)
  
  output$distance_results = renderPlot({
    if (all(is.na(vals$distances))) return()
    
    xvals = vals$plot_every_iterations * (1:vals$number_of_loops)
    plot(xvals, vals$distances, type='o', pch=19, cex=0.7, 
         ylim=c(0, max(vals$distances, na.rm=TRUE)), xlab="iterations", ylab="current tour distance",
         main="Evolution of Current Tour Distance")
  }, height=260)
  
  session$onSessionEnded(function() {
    run_annealing_process$suspend()
    set_cities_randomly$suspend()
  })
})


================================================
FILE: ui.R
================================================
library(shiny)
if (!exists("all_cities")) all_cities = readRDS("data/cities.rds")
if (!exists("usa_cities")) usa_cities = readRDS("data/usa_cities.rds")

shinyUI(fluidPage(
  tags$head(
    tags$link(rel="stylesheet", type="text/css", href="custom_styles.css")
  ),
  
  title = "Traveling Salesman with Simulated Annealing, Shiny, and R",
  
  tags$h2(tags$a(href="/traveling-salesman", "Traveling Salesman", target="_blank")),
  
  plotOutput("map", height="550px"),
  
  fluidRow(
    column(5,
      tags$ol(
        tags$li("Customize the list of cities, based on the world or US map"),
        tags$li("Adjust simulated annealing parameters to taste"),
        tags$li("Click the 'solve' button!")
      )
    ),
    column(3,
      tags$button("SOLVE", id="go_button", class="btn btn-info btn-large action-button shiny-bound-input")
    ),
    column(3,
      HTML("<button id='set_random_cities_2' class='btn btn-large action-button shiny-bound-input'>
              <i class='fa fa-refresh'></i>
              Set Cities Randomly
            </button>")
    ), class="aaa"
  ),
  
  hr(),
  
  fluidRow(
    column(5,
      h4("Choose a map and which cities to tour"),
      selectInput("map_name", NA, c("World", "USA"), "World", width="100px"),
      p("Type below to select individual cities, or", actionButton("set_random_cities", "set randomly", icon=icon("refresh"))),
      selectizeInput("cities", NA, all_cities$full.name, multiple=TRUE, width="100%",
                     options = list(maxItems=30, maxOptions=100, placeholder="Start typing to select some cities...",
                                    selectOnTab=TRUE, openOnFocus=FALSE, hideSelected=TRUE)),
      checkboxInput("label_cities", "Label cities on map?", FALSE)
    ),
    
    column(2,
      h4("Simulated Annealing Parameters"),
      inputPanel(
        numericInput("s_curve_amplitude", "S-curve Amplitude", 4000, min=0, max=10000000),
        numericInput("s_curve_center", "S-curve Center", 0, min=-1000000, max=1000000),
        numericInput("s_curve_width", "S-curve Width", 3000, min=1, max=1000000),
        numericInput("total_iterations", "Number of Iterations to Run", 25000, min=0, max=1000000),
        numericInput("plot_every_iterations", "Draw Map Every N Iterations", 1000, min=1, max=1000000)
      ),
      class="numeric-inputs"
    ),
    
    column(5,
      plotOutput("annealing_schedule", height="260px"),
      plotOutput("distance_results", height="260px")
    )
  )
))


================================================
FILE: www/custom_styles.css
================================================
.recalculating { opacity: 1 !important; }
.numeric-inputs input { width: 75px; }
#go_button, #set_random_cities_2 {
  width: 100%;
  margin-bottom: 10px;
}
hr { margin: 8px 0; }