Optimizing on-demand food delivery with BDI-based multi-agent systems and Monte Carlo tree search scheduling

On-demand food delivery services are a rapidly expanding sector within the logistics industry, yet optimizing delivery routes in real-time remains a significant challenge, particularly in high-demand and complex environments. This gap hinders operational efficiency and customer satisfaction, highlighting the need for advanced decision-making frameworks. In response, we propose a multi-agent system (MAS) using the Belief-Desire-Intention (BDI) framework to enhance delivery efficiency. Our dynamic model simulates interactions between platforms, riders, and shops, utilizing Monte Carlo Tree Search (MCTS) and Insertion Heuristic methodologies to optimize routes. Through simulations of varying complexity, we demonstrate that MCTS outperforms the Insertion Heuristic, especially in complex scenarios, by effectively managing multiple objectives and maintaining high service quality. These results indicate that advanced intention scheduling methods like MCTS can significantly improve real-time decision-making, thereby enhancing both customer satisfaction and operational efficiency in high-demand delivery contexts.