Multi-Criteria Fleet Optimization

Minimizing Cost and Carbon Emissions While Meeting Demand
A project by Hiya Jain, Yash Thakar, Shruti Jain, Kanchan Dabre

Introduction

Large commercial fleets used for logistics and deliveries contribute significantly to global greenhouse gas (GHG) emissions. This project presents a decision-making framework to support optimal fleet decarbonization through multi-objective optimization, balancing:

• 📉 Operational cost
• 🌍 Carbon emissions
• 📦 Demand satisfaction

---

Problem Statement

Objective:
Design a fleet plan that:

• ✅ Minimizes Total Cost of Ownership (TCO)
• 🌿 Minimizes CO₂ emissions
• 📈 Satisfies evolving demand year-over-year (2023–2038)

---

Methodology

Multi-Criteria Optimization Framework

1. Vehicle Data Collection – Extract key attributes (size, fuel, cost, emissions)
2. Vehicle Mapping – Identify feasible vehicles per demand cluster
3. TOPSIS Ranking – Rank options by acquisition cost, operating cost, and emissions
4. NSGA-II Optimization – Generate Pareto-optimal solutions balancing cost and emissions
5. Iterative Planning – Carry forward fleet to future years, re-evaluating based on need
6. Final Output – Determine optimal fleet combination per year

Key Enhancements

• TOPSIS used to bias population initialization in NSGA-II
• Cost and carbon treated as negative criteria
• Existing vehicles reused in future years (zero acquisition cost)

---

Architecture

---

Parallelization

The algorithm is optimized using Python's multiprocessing module:

• Parallelizes size-distance groups across CPU cores
• Avoids GIL bottlenecks
• Achieves ~36.2% reduction in runtime
• No statistically significant difference in results vs sequential

---

Hypothesis Testing & Results

Cost Minimization (50 Runs)

• p-value = 0.00248
• Null Hypothesis Rejected
• TOPSIS-NSGA-II performs significantly better

Emissions Minimization (50 Runs)

• p-value = 0.00303
• Null Hypothesis Rejected
• TOPSIS-NSGA-II again outperforms standard NSGA-II

Parallel vs Sequential

• Cost p-value = 0.0969
• Emissions p-value = 0.5106
• Null Hypothesis Accepted: Parallel execution improves performance without loss in accuracy

---

📊 Performance Comparison

Metric	NSGA-II	TOPSIS-NSGA-II	% Improvement
Total Cost	$477,716,970.18	$342,202,816.90	28.37% ↓
Total CO₂ Emissions (kg)	188,064,097.70	13,444,303.32	36.20% ↓
Runtime (Parallel vs Serial)	Longer	36.2% Faster	✔️

---

📈 Visual Results

---

📁 Repository Structure