Incremental learning-enhanced ensemble surrogate-assisted evolutionary algorithm for lifelong berth allocation and quay crane assignment problems

Abstract

The berth allocation and quay crane assignment problem (BACAP) is a critical challenge in maritime transport, especially in lifelong scenarios that are rarely addressed in the current literature but essential for practical applications. The Lifelong BACAP (LBACAP) presents new challenges, such as the uncertain arrival of vessels, limited resources, and inter-dependencies between vessels. To address these challenges, we propose an incremental learning-enhanced ensemble surrogate-assisted evolutionary algorithm, named IL-ESAEA, with three core designs. (1) The adaptive rolling-horizon strategy divides the LBACAP into consecutive time windows, each corresponding to an interconnected sub-LBACAP. (2) The ensemble surrogate-assisted evolutionary algorithm (ESAEA) approximates the computationally intensive and intricately designed decoding method for optimization, reducing computational costs while maintaining robust search capabilities for solving various BACAPs. (3) The incremental learning mechanism identifies connections between sub-LBACAPs in successive time windows, utilizing historical decisions to guide the optimization effectively. Experimental results demonstrate that IL-ESAEA consistently outperforms state-of-the-art algorithms and provides superior solutions with increased computational efficiency over time. This highlights the strong competitive edge of IL-ESAEA in solving LBACAPs.