Optimizing sugarcane bale logistics operations: Leveraging reinforcement learning and artificial multiple intelligence for dynamic multi-fleet management and multi-period scheduling under machine breakdown constraints

Abstract

This study enhances the bio-circular green economic model within the sugar industry by advancing sustainable practices, notably green harvesting. A significant challenge involves establishing an efficient supply chain for sugarcane bale collection, emphasizing the minimization of idle time and the optimization of travel routes. The objective is to refine the scheduling and routing strategies for specialized machinery in sugarcane bale operations through a heuristic-driven methodology. The incorporation of a reinforcement learning-artificial multiple intelligence system (RL-AMIS) tackles the logistical challenges, particularly in dynamic multifleet scheduling and breakdown management, providing an advanced solution for bale collection. This system combines reinforcement learning (RL) with artificial multiple intelligence system (AMIS) components to enhance profitability. Furthermore, the application of genetic algorithms (GA) and differential evolution (DE) introduces robust enhancement techniques. In support of real-time decision-making for route planners, the study developed sugarcane bale logistics software, BaleLogistics, and corresponding mobile applications based on the RL-AMIS framework. A case study conducted in Thailand demonstrated that the RL-AMIS model significantly outperformed conventional methods, reducing operational costs by 26.1 %, the makespan by 10.54 %, and working time by 6.43 %, while achieving a task completion rate of 96.65 % and decreasing machine downtime by 77.05 %. This research marks a pioneering step in employing technology to optimize sugarcane bale logistics, potentially extending novel and efficient logistic solutions across the agricultural sector.