Pull Request for Hackathon 2020 - JediMasters

JediMasters/DWave Challenge/CDL_Hackathon_2020_FoodBankDeliveryScheduling.ipynb

@@ -0,0 +1,377 @@
+{
+  "nbformat": 4,
+  "nbformat_minor": 0,
+  "metadata": {
+    "colab": {
+      "name": "CDL_Hackathon_2020-JediMasters.ipynb",
+      "provenance": [],
+      "collapsed_sections": [],
+      "authorship_tag": "ABX9TyOVkbtDsT8ddCrTD9WjVH0E",
+      "include_colab_link": true
+    },
+    "kernelspec": {
+      "name": "python3",
+      "display_name": "Python 3"
+    }
+  },
+  "cells": [
+    {
+      "cell_type": "markdown",
+      "metadata": {
+        "id": "view-in-github",
+        "colab_type": "text"
+      },
+      "source": [
+        "<a href=\"https://colab.research.google.com/github/tina-seb/Hackathon2020/blob/master/JediMasters/CDL_Hackathon_2020_FoodBankDeliveryScheduling.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {
+        "id": "7-W1uufjf2Yp",
+        "colab_type": "text"
+      },
+      "source": [
+        "# **Food Bank Delivery Route Scheduler**"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {
+        "id": "ye5iy8A7Z-Pa",
+        "colab_type": "text"
+      },
+      "source": [
+        "**Problem** : Food banks use volunteers on a daily basis to deliver food and groceries to the homebound. Currently they depend on manual scheduling and/or classical software to schedule routes for a variable number of volunteers, which may change during the day.\n",
+        "\n",
+        "**Challenges** : The route maps need to be generated quickly if a volunteer drops off, and also need to be the most cost-efficient and fastest route possible.\n",
+        "\n",
+        "**Proposed Solution** : We implement a variation of the Multiple Travelling Salesman algorithm using DWave Leap Hybrid Solver and QBSolve to generate the most efficient routes for a variable number of volunteers, by minimizing the cost as energy and total mileage.\n",
+        "\n",
+        "**Constraints** : \n",
+        "1) All volunteeers must start and end at the same food bank location.\n",
+        "2) Each volunteer should only visit one home for delivery.\n",
+        "3) If a volunteer does not show up or cancels, another volunteer should be able to pick up his or her route as an extension, without having to reschedule every other route.\n",
+        "4) The cost of each individual route as well as the cost of the entire schedule should be minimized."
+      ]
+    },
+    {
+      "cell_type": "code",
+      "metadata": {
+        "id": "LWSG8rcob1ug",
+        "colab_type": "code",
+        "colab": {}
+      },
+      "source": [
+        "from dwave_qbsolv import QBSolv\n",
+        "from dwave.system import LeapHybridSampler\n",
+        "import numpy as np\n",
+        "import matplotlib.pyplot as plt\n",
+        "import re"
+      ],
+      "execution_count": null,
+      "outputs": []
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {
+        "id": "cRJL6Pohc8_r",
+        "colab_type": "text"
+      },
+      "source": [
+        "The number of houses with homebound clients served by a food bank changes every day. This is currently defined as a variable and can be changed into an input variable. The home addresses and the distance information of each address needs to be provided as input files to the algorithm."
+      ]
+    },
+    {
+      "cell_type": "code",
+      "metadata": {
+        "id": "nGR7Wx-jcm-6",
+        "colab_type": "code",
+        "colab": {}
+      },
+      "source": [
+        "# Number of houses in our food delivery problem\n",
+        "Total_Number_Houses = 48\n",
+        "Number_Volunteers = 8\n",
+        "Number_Routes = (int)(Total_Number_Houses/Number_Volunteers)\n",
+        "\n",
+        "# Tunable parameters. \n",
+        "A = 8500\n",
+        "B = 1\n",
+        "chainstrength = 4500\n",
+        "numruns = 100"
+      ],
+      "execution_count": null,
+      "outputs": []
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {
+        "id": "rAiho6obdpHA",
+        "colab_type": "text"
+      },
+      "source": [
+        "We initialize and construct a distance matrix where the row and column represent a house address and each entry is a distance in metres."
+      ]
+    },
+    {
+      "cell_type": "code",
+      "metadata": {
+        "id": "Iol81rlvd2Wf",
+        "colab_type": "code",
+        "colab": {}
+      },
+      "source": [
+        "## Constructing the distance matrix\n",
+        "D = [[0 for z in range(Total_Number_Houses)] for y in range(Total_Number_Houses)]\n",
+        "\n",
+        "# Input file containing inter-house distances for all counties\n",
+        "fn = \"FoodBankClientDeliveryData.txt\"\n",
+        "\n",
+        "# check that the user has provided input file\n",
+        "try:\n",
+        "  with open(fn, \"r\") as myfile:\n",
+        "    distance_text = myfile.readlines()\n",
+        "    myfile.close()\n",
+        "except IOError:\n",
+        "  print(\"Input inter-house distance file missing\")\n",
+        "  exit(1)\n",
+        "\n",
+        "# Extract the distances from the input file. \n",
+        "for i in distance_text:\n",
+        "  if re.search(\"^between\", i):\n",
+        "    m = re.search(\"^between_(\\d+)_(\\d+) = (\\d+)\", i)\n",
+        "    housea = int(m.group(1))\n",
+        "    houseb = int(m.group(2))\n",
+        "    D[housea][houseb] = D[houseb][housea] = int(m.group(3))\n"
+      ],
+      "execution_count": null,
+      "outputs": []
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {
+        "id": "VKVAwVx3d8X9",
+        "colab_type": "text"
+      },
+      "source": [
+        "We then construct a QUBO (Quadratic Unconstrained Binary Optimization) to feed into the various DWave samplers. We define the objective and constraints based on our preidentified requirements here."
+      ]
+    },
+    {
+      "cell_type": "code",
+      "metadata": {
+        "id": "3etPBbyEeoi-",
+        "colab_type": "code",
+        "colab": {}
+      },
+      "source": [
+        "\n",
+        "# Function to compute index in QUBO for variable \n",
+        "def return_QUBO_Index(a, b):\n",
+        "    return (a)*Total_Number_Houses+(b)\n",
+        "\n",
+        "## Creating the QUBO\n",
+        "# Start with an empty QUBO\n",
+        "Q = {}\n",
+        "for i in range(Total_Number_Houses*Total_Number_Houses):\n",
+        "    for j in range(Total_Number_Houses*Total_Number_Houses):\n",
+        "        Q.update({(i,j): 0})\n",
+        "\n",
+        "# Constraint that each row has exactly one 1, constant = N*A\n",
+        "for v in range(Total_Number_Houses):\n",
+        "    for j in range(Total_Number_Houses):\n",
+        "        Q[(return_QUBO_Index(v,j), return_QUBO_Index(v,j))] += -1*A\n",
+        "        for k in range(j+1, Total_Number_Houses):\n",
+        "            Q[(return_QUBO_Index(v,j), return_QUBO_Index(v,k))] += 2*A\n",
+        "            Q[(return_QUBO_Index(v,k), return_QUBO_Index(v,j))] += 2*A\n",
+        "\n",
+        "# Constraint that each col has exactly one 1\n",
+        "for j in range(Total_Number_Houses):\n",
+        "    for v in range(Total_Number_Houses):\n",
+        "        Q[(return_QUBO_Index(v,j), return_QUBO_Index(v,j))] += -1*A\n",
+        "        for w in range(v+1,Total_Number_Houses):\n",
+        "            Q[(return_QUBO_Index(v,j), return_QUBO_Index(w,j))] += 2*A\n",
+        "            Q[(return_QUBO_Index(w,j), return_QUBO_Index(v,j))] += 2*A\n",
+        "\n",
+        "# Objective that minimizes distance\n",
+        "for u in range(Total_Number_Houses):\n",
+        "    for v in range(Total_Number_Houses):\n",
+        "        if u!=v:\n",
+        "            for j in range(Total_Number_Houses):\n",
+        "                Q[(return_QUBO_Index(u,j), return_QUBO_Index(v,(j+1)%Total_Number_Houses))] += B*D[u][v]\n"
+      ],
+      "execution_count": null,
+      "outputs": []
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {
+        "id": "O2t7kZISevtf",
+        "colab_type": "text"
+      },
+      "source": [
+        "We have two approaches below : A classical solver approach using QBSolv and a Quantum solver approach using Leap Hybrid Sampler. No significant timing differences were found between the two solutions."
+      ]
+    },
+    {
+      "cell_type": "code",
+      "metadata": {
+        "id": "WWSWT04AfRZO",
+        "colab_type": "code",
+        "colab": {}
+      },
+      "source": [
+        "# Run the QUBO using qbsolv (classically solving)\n",
+        "#resp = QBSolv().sample_qubo(Q)\n",
+        "\n",
+        "# Use LeapHybridSampler() for faster QPU access\n",
+        "sampler = LeapHybridSampler()\n",
+        "resp = sampler.sample_qubo(Q)\n",
+        "\n",
+        "# First solution is the lowest energy solution found\n",
+        "sample = next(iter(resp))"
+      ],
+      "execution_count": null,
+      "outputs": []
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {
+        "id": "SGFuyufYfTrS",
+        "colab_type": "text"
+      },
+      "source": [
+        "Display the results obtained as a schedule of route maps for the required number of routes."
+      ]
+    },
+    {
+      "cell_type": "code",
+      "metadata": {
+        "id": "Q0CaOlx0Z2D7",
+        "colab_type": "code",
+        "colab": {}
+      },
+      "source": [
+        "# Display energy for best solution found\n",
+        "print('Energy: ', next(iter(resp.data())).energy)\n",
+        "\n",
+        "# Print route for solution found\n",
+        "route = [-1]*Total_Number_Houses\n",
+        "for node in sample:\n",
+        "    if sample[node]>0:\n",
+        "        j = node%Total_Number_Houses\n",
+        "        v = (node-j)/Total_Number_Houses\n",
+        "        route[j] = int(v)\n",
+        "\n",
+        "# Compute and display total mileage\n",
+        "mileage = 0\n",
+        "for i in range(Total_Number_Houses):\n",
+        "    mileage+=D[route[i]][route[(i+1)%Total_Number_Houses]]\n",
+        "\n",
+        "print('Mileage: ', mileage)\n",
+        "\n",
+        "# Print route:\n",
+        "\n",
+        "houses = [',']*Total_Number_Houses\n",
+        "cities = [',']*Total_Number_Houses\n",
+        "\n",
+        "with open('FoodBankHouseLookup.txt', \"r\") as myfile:\n",
+        "    address_text = myfile.readlines()\n",
+        "    myfile.close()\n",
+        "\n",
+        "for i in address_text:\n",
+        "    index, city, house = i.split(',')\n",
+        "    cities[int(index)] = city.rstrip()\n",
+        "    houses[int(index)] = house.rstrip()\n",
+        "\n",
+        "output = open('Food_Bank_Delivery.route.offline', 'w')\n",
+        "\n",
+        "r_no = 1\n",
+        "for i in range(Total_Number_Houses):\n",
+        "    if(i % Number_Routes == 0):\n",
+        "        output.write('\\n' + 'Route ' + str(r_no) + '\\n')\n",
+        "        r_no = r_no + 1\n",
+        "    output.write(houses[route[i]] + ',' + cities[route[i]] + '\\n')\n",
+        "\n"
+      ],
+      "execution_count": null,
+      "outputs": []
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {
+        "id": "gB9pntwGfqFD",
+        "colab_type": "text"
+      },
+      "source": [
+        "The route schedule output is as follows :\n",
+        "\n",
+        "\n",
+        "Route 1\n",
+        "9980 Brookside Dr.,Columbus OH\n",
+        "417 Theatre Lane,Columbus OH\n",
+        "26 Walnut Drive,Columbus OH\n",
+        "7085 Myers St.,Columbus OH\n",
+        "821 Garfield Rd.,Columbus OH\n",
+        "7297 North Park Road,Columbus OH\n",
+        "\n",
+        "Route 2\n",
+        "16 Park Ave.,Columbus OH\n",
+        "51 Lexington Dr.,Columbus OH\n",
+        "31 York Drive,Columbus OH\n",
+        "349 W. Rockaway Drive,Columbus OH\n",
+        "19 Center Street,Columbus OH\n",
+        "7750 John Drive,Columbus OH\n",
+        "\n",
+        "Route 3\n",
+        "11 High Noon St.,Columbus OH\n",
+        "69 Buckingham St.,Columbus OH\n",
+        "8891 Vine Street,Columbus OH\n",
+        "977 Elizabeth Lane,Columbus OH\n",
+        "78 High Point St.,Columbus OH\n",
+        "44 NW. San Pablo St.,Columbus OH\n",
+        "\n",
+        "Route 4\n",
+        "82 Paris Hill St.,Columbus OH\n",
+        "23 S. Woodside Road,Columbus OH\n",
+        "7655 Catherine Road,Columbus OH\n",
+        "70 Coffee Ave.,Columbus OH\n",
+        "7152 Augusta St.,Columbus OH\n",
+        "5 South Greystone Street,Columbus OH\n",
+        "\n",
+        "Route 5\n",
+        "192 West Albany St.,Columbus OH\n",
+        "7 Marvon Road,Columbus OH\n",
+        "92 South Country Club Rd.,Columbus OH\n",
+        "98 Galvin Street,Columbus OH\n",
+        "3 Hillcrest Court,Columbus OH\n",
+        "9070 Ramblewood Ave.,Columbus OH\n",
+        "\n",
+        "Route 6\n",
+        "8043 N. Glen Ridge Dr.,Columbus OH\n",
+        "7678 Clinton Road,Columbus OH\n",
+        "685 Oakland Court,Columbus OH\n",
+        "737 Pacific Street,Columbus OH\n",
+        "32 Southampton Street,Columbus OH\n",
+        "9697 Bowman Drive,Columbus OH\n",
+        "\n",
+        "Route 7\n",
+        "9304 Brickyard St.,Columbus OH\n",
+        "8858 Canal Lane,Columbus OH\n",
+        "8 Old Jackson Avenue,Columbus OH\n",
+        "51 Columbia Street,Columbus OH\n",
+        "1961 Cleveland Ave,Columbus OH\n",
+        "7798 Brown Drive,Columbus OH\n",
+        "\n",
+        "Route 8\n",
+        "3432 Main St,Columbus OH\n",
+        "23 Oak Meadow Drive,Columbus OH\n",
+        "287 Military Dr.,Columbus OH\n",
+        "1944 Frank Dr,Columbus OH\n",
+        "64 Illinois Lane,Columbus OH\n",
+        "953 South Fieldstone Drive,Columbus OH\n"
+      ]
+    }
+  ]
+}