cgirabawe/nemt-kthouz-prototype

nemt-vehicle-dispatching

This project aims at optimizing the matching between ride requests and cars.

Problem statement

Consider a city with limited boundaries. We have

• —N number of vehicles located at different locations of the city. Each vehicle can be described by their vehicle ID, size, current location (longitude, latitude) and availability times.
• —M number of riders requesting pickups. Each rider can be described by their vehicle ID, number of passengers, pickup address, destination (latitude, longitude) and procedure time.
• —City bounding boxes, static and real-time traffic information across the city

Our ask is to match vehicles to as many riders as possible. Some constraints are to be consider

• —Load size. Car fitting the right load size are to be considered and priority being given to the smaller size and less expensive vehicle
• —Traffic information. Even though the distance between a vehicle and a rider might be short, priority should be given to the vehicle facing less traffic to reach the rider

Solution

The problem is approached as a Capacitated Pickup-Delivery Problem with Time Windows. OSRM and VROOM were used in the backend respectively for distance/duration calculations and routing optimization.

Caveat

[ ] All rides are considered carpoolable. In the future, we need to respect ride criteria being car-poolable or not [ ] All vehicles are considered not having any break. Any day they available, they will be ready to work 8AM to 5PM

Resources

• —Generating pdf files: bbbike.org and geofabrik
• —osrm setup with docker
• —vroom-docker

How to setup

1. 1.First setup OSRM and VROOM
2. 2.Ensure you are in the root directory
3. 3.Run bash install.sh and follow instructions
4. 4.Run the app with gradio app.py