MULTI-AGENT DEEP REINFORCEMENT LEARNING AND HYBRID SIMULATION FOR RESOURCE ALLOCATION IN PORT OPERATIONS

This study develops a hybrid simulation framework integrating multi-agent deep reinforcement learning (MADRL) for port operations. Unlike traditional methods that optimize berths, AGVs, and Yard Blocks separately, our approach treats them as interconnected agents to enable collaborative optimization. The framework combines discrete event and agent-based simulation, allowing agents to iteratively learn optimal policies. Results demonstrate significant reductions in vessel service time, waiting time, and delays, improving overall operational efficiency.

Author dc:creator

• HUANG YING

• Port Operations; Resource Allocation; Multi-Agent Deep Reinforcement Learning (MADRL); Hybrid Simulation; Collaborative Optimization; Operational Effi

Licence

• https://scholarbank.nus.edu.sg/bitstreams/279f6e22-3a42-4e84-9de3-a51f080fa01a/download