An integrated model for coordinating adaptive platoons and parking decision-making based on deep reinforcement learning

Abstract

Improving transportation efficiency is a key challenge in operating automated container terminals (ACTs), particularly in managing yard intersections and optimizing parking resources. However, existing studies often treat these two aspects independently, failing to consider their combined impact on vehicle operation efficiency. To this end, this study proposes a hierarchical control framework, named CAP-PDM, to integrate intersection platoon strategy and parking resource management for Intelligent Autonomous Vehicles (IAVs). The CAP-PDM comprises two layers: (1) the adaptive platoon layer leverages real-time traffic data to determine optimal platoon sizes at intersections, addressing localized congestion and reducing delays; (2) the parking strategy optimization layer achieves dynamic IAV scheduling and task allocation within the horizontal transportation network by considering multiple objectives (i.e., the current efficiency of IAVs and task completion). The Dual Deep Deterministic Policy Gradient (DDPG) algorithm is employed to determine platoon sizes and manage real-time IAV assignments to parking areas. Simulation results demonstrate that compared with other control methods, CAP-PDM demonstrates superior adaptability to varying traffic conditions, minimizes delays, and significantly enhances the operational efficiency of IAVs in ACTs. This study highlights the importance of integrating traffic control with resource optimization to improve the efficiency of automated port operations. The findings provide port managers with innovative insights for optimizing horizontal transportation systems.