dungtran209 / Modelling-and-Analysis-of-a-Vehicle-Routing-Problem-with-Time-Windows-in-Freight-Delivery

A MSc's Dissertation Project which focuses on Vehicle Routing Problem with Time Windows (VRPTW), using both exact method and heuristic approach (General Variable Neighbourhood Search)

Modelling and Analysis of a Vehicle Routing Problem with Time Windows in Freight Delivery

Introduction

A MSc's Dissertation Project which focuses on Vehicle Routing Problem with Time Windows (VRPTW), using both exact method and heuristic approach (General Variable Neighbourhood Search - GVNS).

The exact approach is based on a classical Mix-Integer Programming (MIP) model and solved by IBM ILOG CPLEX 12.9. Both exact and heuristic algorithm are implemented by Python 3.7 and run on a ASUS Intel Core i7 with 1.8 GHz CPU and 8 GB RAM.

Data

Both methodologies are first tested on the famous Solomon (1987) benchmark instances. The data is avalaible on SINTEF.

Then, the methods are applied to solve an logistics problem in a UK-based company.

Methodology Summary

1. Exact Method: Mathematical Formulations

MIP Model: View

2. General Variable Neighbourhood Search (GVNS)

Pseudo-code: View

Flowchart: View

Detailed elements of the GVNS:

Element	Content
Initial Solution Creation	Solomon I1 Heuristic and Clark & Wright Savings Heuristic
Improvement Operator	2-opt, Or-opt, 2-opt*, Relocation, Exchange, CROSS, ICROSS, GENI, 𝜆-interchange
Local Search Process	Variable Neighbourhood Descent (Best-accept strategy)
Stopping Criteria	All neighbours (improvement operators) are explored

Visualization

Best solution (minimum total distance of all tours) found by the GVNS for benchmark instance R206 (100 customers)