Intelligent IoT Based Supply Chain For Fresh Produce: A Hybrid Reinforcement Learning And Optimization Approach

Abstract

Fruits and vegetables form a vital component of the global economy; however, their distribution poses complex logistical challenges due to high perishability, supply fluctuations, strict quality and safety standards, and environmental sensitivity. In this thesis, we propose an adaptive optimization model that accounts for delays, travel time, and temperature variations impacting produce shelf life, and compare its performance against traditional methods such as Robust Optimization (RO), Distributionally Robust Optimization (DRO), and Stochastic Programming (SP).

Our adaptive model significantly outperforms traditional methods by enabling real-time route and temperature adjustments during transit. Empirical results using synthetic and IoT-derived data inspired by a realistic last mile delivery in Toronto show that the adaptive model improves average product shelf life by over 18% and reduces freshness deviation by 80%, with only a marginal increase in travel time. Furthermore, we introduce a Hybrid Model that combines pre-optimized static routes with real-time RL-based corrections. This approach mitigates the limitations of both static and reactive planning by following optimal routes under normal conditions and dynamically overriding them in response to disruptions such as traffic delays or temperature excursions.

Our results demonstrate that the proposed framework retains global efficiency while providing localized adaptability, making it a robust and practical solution for cold-chain logistics. This thesis thus offers a comprehensive, data-driven strategy to enhance sustainability, minimize spoilage, and increase responsiveness in fresh produce supply chains.