This research addresses the optimization of last-mile delivery operations within the Fast-Moving Consumer Goods (FMCG) sector, a domain characterized by high order volumes and strict delivery windows. The primary challenge lies in making real-time, economically efficient decisions regarding vehicle routing and driver schedules while facing significant uncertainty in customer demand and resource availability. To tackle this sequential decision-making problem, the project operates at the intersection of Mathematical Optimization and Machine Learning.
Specifically, it focuses on developing self-learning algorithms, utilizing Approximate Dynamic Programming (ADP), that can dynamically adapt to changing market signals. The goal is to bridge the gap between theoretical reinforcement learning models and practical industry constraints, ultimately producing robust, data-driven methodologies that enhance operational efficiency in stochastic logistics environments.
