VRP Microservice

Vehicle Routing Problem microservice for IDP project at TUM.

Development

Setting up the virtual environment

To work on this project locally, you need to configure a python virtual environment (venv) and install the project dependencies. This virtual environment needs to be activated while working on the project and running python related commands.

It is recommended to use Python version 3.9, as the serverless function works on this version once deployed on Vercel.

• Change your working directory into the /vrpms and create a virtual environment.

  cd /vrpms
  python -m venv venv

• Activate the virtual environment.

  source venv/bin/activate       # if using bash
  source venv/bin/activate.fish  # if using fish

• The project dependencies are specified in the requirements.txt file. To install these pypi packages, run the following:

  pip install -r "requirements.txt"

Folder structure

The project is divided into two important directories, /api and /src.

• The /api directory contains the HTTP endpoints which will be served when a GET, PUT, etc. request is received. No algorithm for the VRP problem should be implemented here. Instead, the necessary functions for calculating the desired response should be called and returned.
• The /src directory contains the algorithms and helper functions.

Importing modules

Python imports can be tricky. To prevent confusion and import issues, it is encouraged to use absolute imports everywhere.

The module src/foo.py can import another module src/utilities/baz.py as follows:

import src.utilities1.baz  # absolute import
from src.utilities1.baz import some_function  # absolute import

import.utilities.baz  # relative import
from .utilities.baz import some_function  # relative import

Secrets

Make sure not to put any API keys and other secrets in the application code, which needs to be placed in the .env file at the top level for local development. These can be set and then accessed as follows:

.env

MY_API_KEY="12345678"

main.py

import os
MY_API_KEY = os.getenv('MY_API_KEY')

However, keep in mind this only works for local development. The environment variables need to be configured at Vercel for deployments.

Alternative Approach for saving keys

In case the user faces problems accessing to the location or duration data saved in the Supabase Database, the following can be done.

If the user has PyCharm IDE, there can be different run profiles generated. The following keys of the Supabase Database can be used to access location and duration data.

SUPABASE_URL = https://pkeygmzuwfucblldmkjn.supabase.co

SUPABASE_KEY = eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJpc3MiOiJzdXBhYmFzZSIsInJlZiI6InBrZXlnbXp1d2Z1Y2JsbGRta2puIiwicm9sZSI6ImFub24iLCJpYXQiOjE2OTc5NzU0MTksImV4cCI6MjAxMzU1MTQxOX0.5TsK_pH0xsMyJWo_XPXt4NhsuS-vW6MAcj575WskZ8s

Note that these values are anonymous access keys and not secrets. Sharing these does not constitute a security risk.

These values must be added following the screenshots provided.

• Please click on the highlighted area to open run configurations list.

• 

• Then click on the edit configurations options.

• 

• Then click on the icon highlighted in the picture.

• 

• Add the provided SUPABASE_URL SUPABASE_KEY key value pairs to the highlighted section.

• 

Deployment

Commit and push the changes to main branch. The serverless function will automatically be deployed in ~20 seconds. You can run the deployed function at https://vrpms-main.vercel.app

Formatting

Code is formatted with black, and it can be set up as pre-commit-hooks via pre-commit. To get started, install the dependencies.

$ pip install -r requirements.dev

Then install the pre-commit hooks:

$ pre-commit install

Format the entire codebase under the root:

$ black --line-length 120 .

You can set up black as a file watcher in PyCharm so that your code is automatically formatted on each save, as documented in the black docs_.

Test

You should install the required packages for testing first.

$ pip install -r requirements.test

How to send POSTMAN queries to Vercel deployment?

Example: Sending a TDVRP Request to Genetic Algorithm endpoint in Vercel Deployment

1. Send to the following address $ https://vrpms-main.vercel.app/api/vrp/ga

   • Explanation of the address is as follows $ https://vrpms-main.vercel.app/<vercel_access_point>/<problem_name>/<algorithm_name>

2. Use the following body as the case-1 example (single core, 48 iterations, 125 population count)

    {
        "solutionName": "GA-TDVRP-MAIN-BRANCH-UNIT-DEMAND-CASE-1",
        "solutionDescription": "IDP-Postman-Request-Example",
        "locationsKey": 4,
        "durationsKey": 3,
        "capacities": [5, 5, 5],
        "startTimes":[0, 0, 0],
        "ignoredCustomers":[26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,
                            41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57],
        "completedCustomers":[],
        "multiThreaded": false,
        "randomPermutationCount": 125,
        "iterationCount": 48,
        "max_k": -1,
        "k_lower_limit": true,
        "auth": "authentication token is only necessary for saving to the DB,
                 in this test it is not necessary. If you would like to get a 
                 token for saving to the DB, run the get_token.py file with your 
                 credentials to get a token"
}

• DISCLAIMER: The Vercel Deployment uses free-tier servers. Thus, heuristic algorithms can be called only with certain hyper-parameter settings. For example running multi core Genetic Algorithm with 5000 population and 144 iteration count is not possible with the Vercel deployment. Because the platform has a time limit for the free-tier servers. Thus, the user has to either buy premium servers or should check the following section called How to run local simulations?.

• Example Request Screenshot

List of available endpoints for algorithms:

• https://vrpms-main.vercel.app/api/vrp/ga
• https://vrpms-main.vercel.app/api/tsp/ga
• https://vrpms-main.vercel.app/api/vrp/aco
• https://vrpms-main.vercel.app/api/tsp/aco
• https://vrpms-main.vercel.app/api/vrp/sa
• https://vrpms-main.vercel.app/api/tsp/sa
• https://vrpms-main.vercel.app/api/vrp/bf
• https://vrpms-main.vercel.app/api/tsp/bf

How to run local benchmark tests?

vrpms/test/test_ga_tdvrp is an example of the test script folder. Basically the user can see how we send parameters to our methods. This allows the user to play around with the parameters.

How to run local simulations?

vrpms/src/scenarios/scenario.py can be used with different paramters to run local simulations.

The user can change the inputs of the simulation by changing the inputs of the "run" method.

• changing vrp_algo_params_path changes the vrp algorithm used
• changing tsp_algo_params_path changes the tsp algorithm used
• for example setting vrp_algo_params_path = ../../data/scenarios/vrp/config_vrp_ga_1.json would call the Genetic Algorithm with multi core option enabled
• the detailed options available can be seen under vrpms/data/scenarios