Aniket Sharma

AI powered Routing Tool

As an Artificial Intelligence intern at Kinematic Foodtech, I worked on a Routing Tool to manage the delivery fleet and create routes to reduce the company's logistics costs.

Tech Stack used:

The basic structure of the tool can be divided into two NP-hard problems:

Capacitated Vehicle Routing Problem with Time Windows and Multiple Depots

Minimization Objective: The tools provides option to use two minimization objectives -

Cost of a vehicle is divided into:

Different Vehicles have different cost.

Capacity Constraints:

Time Window Constraints:

Multiple Depot: There are two types of depots -

Note: There is a start time for depots which tells the earlier any delivery riders can leave from there.

Time spent at delivery location:

Facility Location and Allocation Problem

A modified version of k-mean clustering algorithm which uses the distances provided by the Openrouteservice API instead of Euclidean distance for creating facilities/hubs. The position of static hubs is also taken into account while creating the dynamic hubs.

The number of facilities to create is determined by using an automated version of elbow method as described in this article.

The generated hubs are then shown to the operations team to confirm or slightly modify the positions of the dynamic hubs as per their need before they start the route generation.