Scholarlink

Addressing the Three-Dimensional Loading Capacitated Vehicle Routing Problem (3L-CVRP) within the increasingly demanding context of last-mile distribution requires tackling a complex multi-objective optimization task that balances routing decisions with intricate spatial constraints. Given its NP-hard nature, metaheuristic algorithms have become the standard solution approach; however, their performance remains highly sensitive to static parameter configurations that often fail to adapt to the dynamic nature of the search process. This paper investigates the integration of Naive Behavior Cloning (NBC) as an adaptive parameter control mechanism within the Sparrow Search Algorithm (SSA) to solve the multi-objective 3L-CVRP. Unlike traditional methods that rely on offline tuning or historical execution data, the proposed SSA+NBC approach dynamically adjusts algorithm parameters online based on the real-time search state. To validate this method, a comprehensive experimental study was conducted using a heterogeneous set of synthetic 3L-CVRP instances categorized into low, medium, and high complexity scenarios. The adaptive SSA+NBC configuration was compared with a static SSA baseline through multiple performance indicators, including convergence reliability, solution quality, and computational efficiency, with statistical significance verified by the Mann-Whitney U test. The results demonstrate that our proposal significantly enhances execution efficiency, reducing computation time by 45–66% ( \(p < 0.001\) ), while preserving comparable solution quality across all complexity levels.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Solving the 3L-CVRP via Sparrow Search with NBC-Based Adaptive Parameter Control

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Sandy Iturra,
  • Rodrigo Olivares

摘要

Addressing the Three-Dimensional Loading Capacitated Vehicle Routing Problem (3L-CVRP) within the increasingly demanding context of last-mile distribution requires tackling a complex multi-objective optimization task that balances routing decisions with intricate spatial constraints. Given its NP-hard nature, metaheuristic algorithms have become the standard solution approach; however, their performance remains highly sensitive to static parameter configurations that often fail to adapt to the dynamic nature of the search process. This paper investigates the integration of Naive Behavior Cloning (NBC) as an adaptive parameter control mechanism within the Sparrow Search Algorithm (SSA) to solve the multi-objective 3L-CVRP. Unlike traditional methods that rely on offline tuning or historical execution data, the proposed SSA+NBC approach dynamically adjusts algorithm parameters online based on the real-time search state. To validate this method, a comprehensive experimental study was conducted using a heterogeneous set of synthetic 3L-CVRP instances categorized into low, medium, and high complexity scenarios. The adaptive SSA+NBC configuration was compared with a static SSA baseline through multiple performance indicators, including convergence reliability, solution quality, and computational efficiency, with statistical significance verified by the Mann-Whitney U test. The results demonstrate that our proposal significantly enhances execution efficiency, reducing computation time by 45–66% ( \(p < 0.001\) ), while preserving comparable solution quality across all complexity levels.