Predictive Movement

This is the source code for the collaborative open source project Predictive Movement.

About

Predictive Movement aims to create a digital platform that will act as a collaborative hub for the transports of people and goods with the help of artificial intelligence (AI). In its first phase, the project will solve the societal challenge of parcel deliveries in rural areas. Further on, the project aims to deal with optimizing availability and accessibility in areas ranging from urban to rural.

Predictive Movement is a project financed by, among others, Sweden’s Innovation Agency (Vinnova) and the Swedish Transport Administration. The project includes one Swedish region, four municipalities, a university, authorities as well as key actors within digitization and traffic/logistics. An important driving force behind the project is to combat climate change and to reduce emissions caused by road transports.

Navigating the code

The source code for the project is contained in this mono-repo. Within the packages folder you will find all included packages. Here is a summary of the components:

• Engine The main logic written in Elixir
• Driver interface Bot in Telegram for communicating with drivers
• Booking interface Bot in Telegram for communicating with bookers of transport
• Engine UI Main UI for visualising current bookings and cars
• Engine Server Node-server backend for Engine UI
• Route Optimization Jsprit A fork of Jsprit; used to help calculating the best vehicle to booking matches.
• Vehicle Offer Re-routes vehicle offers to the queue
• Auto Accept Offer Auto accepts offers for vehicles created in the Engine UI
• Booking-dispatcher Auto accepts offers for vehicles created in the Engine UI
• Signing UI UI for signing a package on delivery

Running the project

Using kubernetes

Secrets

The following kubernetes secrets are used:

DRIVER_TOKEN - Driver Telegram bot token. Used by driver-interace
GOOGLE_TOKEN - Used by driver-interace
POSTNORD_KEY - Used by engine-server to get information from Postnord API
MINIO_ROOT_PASSWORD - Used by minio and engine-server
POSTGRES_PASSWORD - Used by postgres, engine and postgres-backup

Deployment

To deploy the dependencies of the stack (usually done once and it's DBs) to your Kubernetes cluster, use Skaffold:

skaffold -f skaffold-dependencies.yaml run

To deploy the relevant packages to your Kubernetes cluster, use Skaffold:

skaffold run --tail

To debug run:

skaffold dev

Before you start running the project

Start by exporting port 9200

kubectl port-forward elasticsearch-0 9200:9200 --namespace pelias

Running Predictive Movement locally

You will need the following API keys POSTNORD_KEY - for engine-server REACT_APP_MAPBOX_ACCESS_TOKEN - for engine-ui GOOGLE_API_TOKEN - for driver-interface TELEGRAM_BOT_TOKEN - for driver-interface

Set environment variables

create .env-file in packages/driver-interface/.env with

GOOGLE_API_TOKEN=<FROM LASTPASS>
BOT_TOKEN=<FROM LASTPASS> / or create your own bot in telegram

create .env-file in packages/engine-ui/.env with

REACT_APP_MAPBOX_ACCESS_TOKEN=<FROM LASTPASS>

Start the services

docker-compose up

Run migrations and start the engine

cd packages/engine_umbrella/
mix deps.get
mix setup_dev
iex -S mix

Running tests

Make sure you've started the dependencies with docker-compose up -d then run

cd packages/engine_umbrella/
mix deps.get
mix setupTestDatabase
mix test

Helper/utility functions for populating the state

The umbrella project has an application "message_generator" which is used to create rabbitMQ messages for producing transports and bookings. First start the umbrella project in an elixir shell

cd packages/engine_umbrella/
iex -S mix

Then the Generator module is available inside the shell.

add_booking/0      add_booking/1      add_transport/0    add_transport/1

The argument can be a map containing properties or an atom for creating a generic one close to a city, (:stockholm, :gothenburg, :ljusdal) or in the case of the add_transport; a keyword list can also be used containing :phone, i.e: Generator.add_transport(phone: "0735333")

Design mockup: Figma

---

Release

You will need:

• docker installed
• kubectl installed
• skaffold installed
• kustomize installed
• login with your Docker account
• access to Iteam Kubernetes cluster
• mapbox access token from Predictivemovement LastPass folder

To deploy the dependencies of the stack (usually done once and it's DBs) to your Kubernetes cluster, use Skaffold:

skaffold -f skaffold-dependencies.yaml run --profile prod

Set environment variables that are used by Docker at build time (for the UI) and run the skaffold command with a profile:

export REACT_APP_MAPBOX_ACCESS_TOKEN=<FROM LASTPASS>
export REACT_APP_ENGINE_SERVER=https://engine-server.iteamdev.io
skaffold run --profile prod

---

Data backups

We use postgres-backup

To restore a backup exec into the postgres-backup pod

kubectl exec -it postgres-backup /bin/bash

/restore.sh /backup/latest.psql.gz # or choose a different backup you want

---

Vocabulary

Transport (previously vehicle){
    id
    busy
    activities
    booking_ids
    metadata
    start_address
    end_address
    earliest_start
    latest_end
    profile
    capacity
}

Booking {
    id
    pickup
    delivery
    assigned_to
    external_id
    events
    metadata
    size
    requires_transport_id
}

Plan

Route