Multi-objective optimization for dynamic logistics scheduling based on hierarchical deep reinforcement learning.

Area of Science:

• Operations Research
• Artificial Intelligence
• Logistics Management

Background:

• Modern logistics scheduling faces challenges due to complexity, uncertainty, and competing objectives.
• Traditional optimization and reinforcement learning methods are insufficient for dynamic, multi-objective logistics problems.

Purpose of the Study:

• To propose a novel hierarchical deep reinforcement learning (DRL) framework for multi-objective optimization of dynamic logistics scheduling.
• To address limitations of existing methods in handling complexity, uncertainty, and competing goals.

Main Methods:

• Implemented a two-level hierarchical DRL architecture with high-level strategic planning and low-level tactical execution networks.
• Developed a Pareto-optimal reward mechanism with adaptive weighting to balance time, cost, and service quality.
• Utilized temporal abstraction to improve sample efficiency and handle sparse rewards.

Main Results:

• Achieved significant improvements: 18.4% in service quality, 15.2% reduction in order fulfillment time, and 7.8% decrease in operational costs.
• Demonstrated superior performance compared to state-of-the-art methods on diverse logistics datasets.
• Generated high-quality Pareto-optimal solutions balancing competing objectives effectively.

Conclusions:

• The proposed hierarchical DRL framework offers a robust solution for multi-objective dynamic logistics scheduling.
• The framework excels in dynamic environments, providing strategic coherence and tactical adaptability with near real-time decision-making.
• This approach effectively balances competing objectives, improving overall logistics operations.