Carbon-Efficient Scheduling in Fresh Food Supply Chains with a Time-Window-Constrained Deep Reinforcement Learning Model

Abstract

Intelligent Transportation Systems (ITSs) leverage Internet of Things (IoT) technology to facilitate smart interconnectivity among vehicles, infrastructure, and users, thereby optimizing traffic flow. This paper constructs an optimization model for the fresh food supply chain distribution route of fresh products, considering factors such as carbon emissions, time windows, and cooling costs. By calculating carbon emission costs through carbon taxes, the model aims to minimize distribution costs. With a graph attention network structure adopted to describe node locations, accessible paths, and data with collection windows for path planning, it integrates to solve for the optimal distribution routes, taking into account carbon emissions and cooling costs under varying temperatures. Extensive simulation experiments and comparative analyses demonstrate that the proposed time-window-constrained reinforcement learning model provides effective decision-making information for optimizing fresh product fresh food supply chain transportation and distribution, controlling logistics costs, and reducing carbon emissions.