A discrete collaborative swarm optimizer for resource scheduling problem in mobile cellular networks
A discrete collaborative swarm optimizer for resource scheduling problem in mobile cellular networks
复制标题
DOI：
10.1007/s00521-021-05803-3
复制
发表时间：
2021-03
影响因子：
6
通讯作者：
Bei Dong;Yuping Su;Yun Zhou;Shi Cheng;Xiaojun Wu
中科院分区：
文献类型：
--
作者：
Bei Dong;Yuping Su;Yun Zhou;Shi Cheng;Xiaojun Wu
本研究提出了一种离散协作群优化器（DCCSO）来解决移动蜂窝网络中的资源调度问题，其目的是在不违反干扰约束的情况下采用最小无线带宽来满足每个小区的各种信道需求。当前的许多算法在处理简单问题时可以提供令人满意的解决方案，但由于资源有限，一些复杂问题仍然需要高效的调度方案。该算法受到竞争群体优化器的启发，其在连续优化问题上的优越性已得到理论和验证的证明。针对资源调度问题的特点，设计了广义顺序学习机制，通过学习获胜者的顺序知识来更新失败者粒子的信息。此外，由于原始解空间的探索能力和覆盖速度退化，搜索过程中会产生大量无效解。为此，通过帮助邻域搜索变换后的解决方案空间中的特定问题信息，提出了集成自学习策略。所提出的 DCCSO 的有效性在一组现实问题上得到了证明，实验结果表明所提出的算法在大多数问题上表现出优于或至少与其他最先进算法相当的性能。
This study proposes a discrete collaborative swarm optimizer (DCCSO) for solving the resource scheduling problem in mobile cellular networks, which aims to employ minimum wireless bandwidth to meet various channel demands from each cell without violation of interference constraint. Many current algorithms can provide satisfactory solutions in dealing with simple problems, while some complex problems still need efficient scheduling schema, due to the limited resources. The proposed algorithm is inspired by the competitive swarm optimizer, whose superiority on continuous optimization problems has been proven by theory and verification. With the characteristics of the resource scheduling problem, the generalized order learning mechanism is designed, which updates the information of the loser particles by learning the sequential knowledge of the winners. Besides, plenty of invalid solutions will generate during the searching process in the original solution space degeneration of the exploration capability and coverage speed. To that end, an ensemble self-learning strategy is arisen by helping the neighborhood search by problem-specific information in the transformed solution space. The effectiveness of the proposed DCCSO is demonstrated on a set of real-world problems, and the experimental results show that the proposed algorithm exhibits better than or at least comparable performance to other state-of-the-art algorithms on most problems.