基于弹性碰撞优化算法的传感云资源调度

针对当前智能优化算法普遍存在收敛精度不高、容易“早熟”的缺陷，提出全新的智能优化算法—弹性碰撞优化（ECO）算法.算法基于弹性碰撞物理学现象，通过模拟碰撞过程中物理属性相互影响的变化过程，抽象出“与种群最优碰撞”、“与自身历史最优碰撞”和“随机碰撞”3种粒子更新机制.为了有效提升复杂高维优化问题的寻优能力，设计自适应核模糊C-均值聚类（AKFCM）算法，利用AKFCM对ECO种群进行聚类分析，通过迭代比对策略实现种群自动最佳聚类划分，确保粒子学习对象的合理性与多样性.种群样本多样性定量分析表明ECO在运算后期具有较好的种群多样性.将ECO应用于传感云资源调度问题，为了满足传感云系统管理多样性需求，构建多目标优化传感云资源调度模型，设计符合调度问题的ECO粒子编码方式，实现传感云资源高效率调度优化.多维复杂测试函数以及传感云资源调度实例仿真结果表明，ECO具有较高的收敛精度和成功率，有效降低了传感云资源调度的能耗和任务长度.
Abstract: Aiming at the low convergence accuracy and precocity of current intelligent optimization algorithms, a new intelligent optimization algorithm, elastic collision optimization (ECO) algorithm was proposed. Considering the elastic collision physics phenomenon, the changes of physical properties of objects during collision process were simulated, and three kinds of particle renewal mechanism were proposed, namely, collision with optimal individuals, collision with history optimal, and random collision. In order to effectively improve the optimization ability of complex and high dimensional optimization problems, an adaptive kernel fuzzy C-means (AKFCM) algorithm was designed and the ECO population was analyzed by AKFCM. By using the iterative comparison method, the automatic optimal clustering of the swarm was realized, and the rationality and diversity of the particle learning object were ensured. Quantitative analysis of population sample diversity showed that ECO still had good population diversity at the later stage. ECO was applied to the sensing cloud resource scheduling problem. The resource scheduling model of sensing cloud based on multi-objective optimization was constructed for the diversity management of sensing cloud systems, and the ECO particle coding was designed for the scheduling problem, which helped realize efficient scheduling and optimization of sensing cloud resources. The simulation results of multidimensional complex test function and sensor cloud resource scheduling show that ECO has higher convergence accuracy and success rate, and effectively reduces the energy consumption and task length of sensing cloud resource scheduling.