摘要
针对现行自动导引车(AGV)系统研究中过分依赖匀速运动和忽略通信时长等模型理想化假设条件及转弯惩罚和拥堵惩罚因子等仿真修正参数的缺点,提出一种面向真实场景的自适应集群调度策略.首先分析了AGV集群系统的地理特征并得到运动约束,然后对于现行的A*算法进行改进,实现加速计算.针对遗传算法适应性有待改善的问题,在融合A*算法的基础上提出了基于动态适应度函数的改进遗传算法,所提出的方法不再须要针对新的环境修改遗传算法结构,只须将约束添加至A*算法工作的地图中.给出了双向A*算法下的多AGV调度的实验结果,分析表明:本研究的双向A*算法在路径求解速度上有明显的优势,且能够适应不同数量的障碍,为AGV集群调度方法在真实作业场景中的开发应用提供了参考.
Aiming at the shortcomings of the ideal model assumptions such as over-reliance on uniform motion and ignoring communication time as well as simulation modification parameters such as turning penalty and congestion penalty factors in the existing researches of automatic guided vehicle(AGV)system,an adaptive cluster scheduling strategy for real scene was proposed.First,the geographical characteristics of the AGV cluster system were analyzed,and the motion constraints were obtained.Then,the existing A*algorithm was improved to achieve accelerated calculation.For strengthening the adaptability of the genetic algorithm,an improved genetic algorithm based on dynamic fitness function was proposed based on the fusion of A*algorithm.The proposed method does not need to modify the genetic algorithm structure for the new environment but only needs to add constraints to the map where A*algorithm works.Experimental results of multi-AGV scheduling with the bidirectional A*algorithm were given,and analysis show that the proposed bidirectional A*algorithm has apparent advantages in path-solving speed and can adapt to different numbers of obstacle,which could provide an alternative reference for developing and applying the AGV cluster scheduling method in real-world scenarios.
作者
郭鹏
汪世杰
周士祺
史海超
GUO Peng;WANG Shijie;ZHOU Shiqi;SHI Haichao(School of Mechanical Engineering,Southwest Jiaotong University,Chengdu 610031,China)
出处
《华中科技大学学报(自然科学版)》
EI
CAS
CSCD
北大核心
2022年第5期123-129,共7页
Journal of Huazhong University of Science and Technology(Natural Science Edition)
基金
国家重点研发计划资助项目(2020YFB1712200).
关键词
智慧仓储
自动导引车(AGV)
并行加速
遗传算法
A*算法
smart warehousing
automatic guided vehicle(AGV)
parallel acceleration
genetic algorithm
A*algorithm