COE Dept. - MS Thesis Defense - Mr. Shihab Hasan

ANNOUNCEMENT-COE MS THESIS

Mr. Shihab Hasan, a Full-Time Computer Networks MS Student, will defend his MS Thesis. His thesis title is “Optimizing Last-Mile Delivery with Hybrid Truck-Drone Systems: A Predictive Traffic MINLP Framework”. His thesis advisor is “Dr. Tarek Sheltami, Professor, COE Department”. You are cordially invited to attend.

Date: Thursday, December 12, 2024 
Time: 11:00 AM to 12:30 PM
Location: Building 76 Room 1238 

Abstract:

The rapid expansion of e-commerce has heightened the need for efficient last-mile delivery solutions, prompting the integration of drones into traditional truck-based logistics. However, optimizing hybrid truck-drone delivery systems poses challenges due to dynamic traffic, drone endurance limits, and coordination requirements. Existing approaches often struggle with computational complexity and lack real-time traffic dynamics. To address these issues, we propose a Predictive Time-Dependent Vehicle Routing Problem with Drones (PTD-VRP-D) framework. Our approach integrates predictive traffic modeling using an XGBoost-based model to forecast truck speeds using real-time traffic data, enabling dynamic adjustment of truck travel times and route optimization under fluctuating conditions. We develop a Mixed-Integer Nonlinear Programming (MINLP) model to optimize both trucks stop locations within delivery clusters and the truck route to enhance drone deployment efficiency. Additionally, we introduce an adaptive clustering technique that dynamically adjusts delivery clusters based on geographic proximity and drone feasibility, along with dynamic drone scheduling, to minimize idle time and optimize fleet utilization. Computational experiments demonstrate that our integrated framework significantly improves total delivery time, route accuracy, and operational efficiency compared to traditional models, especially under variable traffic scenarios. Specifically, it achieves up to a 66% decrease in total delivery time over traditional truck-only delivery methods and up to a 10% improvement over static clustering approaches. Our scalable and adaptive approach addresses key gaps in existing research and offers practical solutions for enhancing last-mile delivery performance in urban logistics.

All faculty, researchers and graduate students are invited to attend.