基于改进NSGA-Ⅲ算法的多AGV任务分配
陈永祥 , 王建平 , 邱子麒 , 代周杰 , 孙杰
盐城工学院学报(自然科学版) ›› 2025, Vol. 38 ›› Issue (4) : 60 -69.
基于改进NSGA-Ⅲ算法的多AGV任务分配
Multi-AGV Path Planning Based on an Improved NSGA-Ⅲ Algorithm
针对仓库多AGV协同调度的拓扑图法无法考虑行驶冲突,栅格图法需对各AGV行驶路径单独规划,从而浪费算力的问题,提出了一种在栅格地图中采用双通道的建模方法和以冲突路口建立拓扑点映射的方法进行路径规划,无需使用单车寻路算法。在算法方面,针对迭代算法过程中广度和深度不能在前后期合理地兼顾,优化目标单一的问题,提出了一种基于NSGA-Ⅲ的多目标优化算法;将算法中的遗传和交叉概率调整为自适应,交叉改为分段交叉,并保留优秀父代片段,从而提高算法的收敛速度;设置考虑AGV系统的能耗、运行时间、负载标准差3个目标函数,并在栅格地图中多AGV 调度问题上进行了仿真实验。结果表明,提出的NSGA-Ⅲ改进算法具有更好的优化结果,收敛代数分布更小;3个目标函数平均收敛速度相比改进前分别提升了26.7%、33.3%、23.5%。
To address the limitations of conventional topological mapping methods for multi-AGV collaborative scheduling in warehouses—particularly the inability to resolve travel conflicts and the computational overhead of separate path planning for each AGV in grid-based approaches—we propose an innovative solution. This method employs dual-channel modeling in grid maps and establishes topological point mappings at conflict intersections for path planning, eliminating the need for single-agent routing algorithms. Algorithmically, we overcome the challenges of balancing breadth and depth during iterative processes and single-objective optimization by introducing a NSGA-III-based multi-objective optimization algorithm. The algorithm incorporates adaptive genetic and crossover probabilities,adopts a segmented crossover scheme, and retains superior parent segments to enhance the algorithm's convergence speed. Three objective functions were defined: energy consumption, operational time, and load standard deviation. Simulation experiments on multi-AGV scheduling in grid maps demonstrated that the improved NSGA-III algorithm achieves better optimization results with smaller convergence algebra distributions. The average convergence speeds of the three objective functions improved by 26.7%,33.3%, and 23.5% respectively compared to pre-improvement benchmarks.
Algorithm1:NSGA-Ⅲ
Input:种群大小N,参考点集R,最大进化代数T
Output:近似Pareto最优解集
1: 初始化种群P0,生成参考点集R
2: 计算P0的参考点关联,得到Cavg(0),Cmax=Cavg(0)
3: for t = 1 to T, do
4: 选择父代个体,基于Pc(t)和Pm(t)通过分段交叉和
多项式变异生成子代Q
5: 合并种群Rt= Pt ∪ Qt
6: 对Rt进行非支配排序,归一化目标空间,关联参考
点Cavg(t)
7: 计算参考点拥挤度Cavg(t),Cmax=max(Cmax,Cavg(t))
8: 多样性比率计算d(t)=Cavg(t)/Cmax
9: 调整概率:
Pc(t)=Pcmax-(Pcmax-Pcmin)*(1-d(t))
Pm(t)=Pmmin+(Pmmax-Pmmin)*(1-d(t))
10: 选择新一代Pt+1
11: end for
| [1] |
张中伟, 高增恩, 王菁锐, |
| [2] |
杨凯. AGV发展现状及关键技术综述[J]. 机电工程技术, 2025, 54(13): 24-30. |
| [3] |
颜伟, 黄冠鹏, 高玉萍, |
| [4] |
刘燕丽. 基于冲突的NP难问题完备算法的研究[D]. 武汉: 华中科技大学, 2019. |
| [5] |
王健. 危险化学品车辆路径问题及算法研究[D]. 北京: 北京化工大学, 2016. |
| [6] |
徐晗, 金隼, 罗磊, |
| [7] |
王占山, 陈晓华, 赵瑞, |
| [8] |
耿焕同, 戴中斌, 王天雷, |
| [9] |
|
| [10] |
|
| [11] |
陈先亮, 黄元君, 范勤勤. 基于多模态多目标进化算法的无人机三维路径规划[J]. 火力与指挥控制, 2023, 48(11): 32-39. |
| [12] |
毕晓君, 王朝. 一种基于参考点约束支配的NSGA-Ⅲ算法[J]. 控制与决策, 2019, 34(2): 369-376. |
安徽省重点实验室开放基金项目(QCKJJ202505)
芜湖市重点研发计划项目(2023ly05)
/
| 〈 |
|
〉 |