Delivering Data: A Real-World Dataset for Last-Mile Delivery Optimization

The dataset captures approximately one month of pharmaceutical delivery operations from a 3PL provider. It represents a Capacitated Vehicle Routing Problem with Time Windows (CVRPTW), incorporating real-world constraints like vehicle capacity, delivery time windows, and operational limits. With nine daily instances covering two delivery days per week, it reflects four to five weeks of routing challenges, providing a benchmark for optimizing last-mile pharmaceutical logistics. 

The dataset  is divided into two main folders:

1. Time and Distance Matrices:

This folder contains travel distances and estimated travel times between delivery nodes for each of the nine days. To maintain confidentiality, exact node locations are not provided.

Distance Matrices: Provide the pairwise distances (in kilometers) between delivery locations

Time Matrices: Represent travel durations (in minutes) under three different traffic conditions:

Optimistic: Best-case scenario with minimal traffic 

Most Likely: Expected travel time 

Pessimistic: Worst-case scenario with heavy traffic delays

Each dataset version (1-9) has four corresponding Excel files:

distance_matrix_X.xlsx

time_matrix_mostlikely_X.xlsx

time_matrix_optimistic_X.xlsx

time_matrix_pessimistic_X.xlsx

where X represents the dataset number (1-9).

2. Order Characteristics:

This folder contains detailed information about daily delivery requests. It includes an Excel file (orders.xlsx) with nine sheets, each corresponding to a specific day's delivery requests. Each order is defined by the following attributes:

NODE_ID: Unique identifier for each delivery request

WEIGHT: Weight of the shipment (kg)

VOLUME: Volume of the shipment (cubic meters)

SERVICE_TIME: Time required to complete the delivery (minutes)

EAT: Earliest Arrival Time allowed for delivery

LAT: Latest Arrival Time allowed for delivery

TIME_WINDOW: Allowed delivery time window, defined by the EAT and LAT