Abstract: This work deals with a multiple-depot cold chain logistics distribution optimization problem with considerations of hard time windows. Firstly, the problem domain is presented and the programming model is formulated to minimize the total cost. Then, an algorithm named ESCA is developed for solution methodology. By virtue of the problem nature, a novel constructive encoding and decoding method, as well as an evaluation function, are designed to adjust SCA to the considered problem. Meanwhile, the opposition-based learning method is embedded into SCA to improve performance of initial solutions. To balance exploration and exploitation abilities, a hybrid individual update method is designed by combing the multi-population mechanism, the non-linear transition and the random perturbation. In addition, a discrete neighbor search method is used to avoid search stagnation. Finally, experiments on the case study and the algorithm comparison analysis are concocted to validate the efficiency of  ESCA.

Key words: optimization, scheduling, sine-cosine algorithm, hybridizations, neighbor search

摘要： 以冷链物流为对象，研究了一类考虑多中心联合配送和硬时间窗约束的调度问题。基于问题描述建立了以最小化总成本为目标的数学模型。提出了改进正余弦算法（enhanced sine-cosine algorithm，ESCA）以获取当前问题的满意解。结合问题特征创建了融合构造式规则的编解码方法，并辅以个体评估方法实现模型与正余弦算法（sine-cosine algorithm，SCA）的适配。同时，将反向学习机制嵌入ESCA的初始化流程，旨在提升初始解的性能。在种群进化方面，构建了融合双种群机制、非线性参数调节和随机扰动的混合进化机制以平衡寻优过程的全局探索和局部挖掘行为，并通过离散邻域搜索方法避免搜索停滞。开展了案例研究和算法对比实验，结果验证了ESCA算法的良好性能。

关键词: 优化, 调度, 正余弦算法, 混合, 邻域搜索