A multimodal deep reinforcement learning approach for IoT-driven adaptive scheduling and robustness optimization in...

Area of Science:

• Logistics and Supply Chain Management
• Artificial Intelligence
• Operations Research

Background:

• Global logistics networks face challenges like demand volatility and disruptions.
• Integrating diverse data sources (IoT, historical) is crucial for effective management.
• Existing methods struggle with dynamic scheduling and robust optimization.

Purpose of the Study:

• To develop an AI-driven framework for adaptive scheduling and robustness optimization in global logistics.
• To integrate multimodal deep reinforcement learning with Internet of Things (IoT) technologies.
• To enhance supply chain resilience and operational efficiency.

Main Methods:

• A multimodal data fusion mechanism synthesizes heterogeneous IoT sensor data, historical records, and contextual information.
• An adaptive deep reinforcement learning architecture generates dynamic scheduling policies.
• A multi-objective robust optimization method balances operational efficiency with system resilience.

Main Results:

• The proposed framework achieved an 18.7% reduction in operational costs and a 12.4% improvement in service levels.
• Demonstrated significantly enhanced robustness, maintaining 83% performance stability during complex disruptions.
• Outperformed traditional methods in real-world logistics datasets.

Conclusions:

• The AI framework effectively addresses global logistics challenges, including demand volatility and disruptions.
• Multimodal learning and robust optimization techniques improve supply chain performance.
• The approach offers a feasible and practical solution for AI-driven logistics operations management.