一种复杂系统仿真研究的新模式SIGD

doi:10.16182/j.issn1004731x.joss.24-0472

摘要/Abstract

摘要：

系统复杂性主要体现在种类和数量众多的组分和极其复杂的相互作用关系。结合目前人工智能的发展趋势，分析思考了对于仿真学科研究带来的思维方式变化，形成了对于仿真智能内涵和研究范畴的理解。提出一种复杂系统仿真研究的新模式：“以决策为愿景的仿真智能(simulation intelligence based generating decisions，SIGD)”。在SIGD研究模式中，仿真学科中的相似性原理、建模方式以及决策引导模式都与传统仿真中的定位有一定的区别。在这种理念指导下，面向连接的仿真世界观成为一种刻画复杂系统的底层逻辑。以主体、连接和结构为基本元素，采用先自顶向下和自底向上相融合的方式编织以神经网络模型为计算单元的“网真”系统。在该系统中，采用代理建模的方式实现从机理模型到神经网络模型的模型变换，并且基于物理信息神经网络(PINN)和图信息神经网络(GINN)表达整个系统。给出了SIGD的技术实现框架：网真神CE平台作为复杂系统仿真新模式的解决方案，以人流仿真为例，探索了“网真”系统的构建过程。

关键词: 复杂系统, 仿真智能, SIGD, PINN, 代理建模

Abstract:

The complexity of the system is mainly reflected in the numerous components and extremely complex interactions. Combined with the current trend of artificial intelligence development, this paper analyzes and considers the changes in thinking mode brought by simulation discipline research, and forms an understanding of the connotation and research scope of simulation intelligence. A new pattern for complex system simulation research is proposed: "simulation intelligence based generating decisions (SIGD)". In the SIGD pattern, the similar principles, modeling methods, and decision-guiding modes in simulation disciplines are different from those in traditional simulation. Under the guidance of this concept, a connection-oriented simulation worldview has become a fundamental logic to characterize complex systems. Taking agents, connections, and structures as basic elements, a "neuro-network model" system with neural network models as computing units is woven using a top-down and bottom-up fusion approach. In this system, the model transformation from mechanism models to neural network models is realized using surrogate model and the entire system is expressed based on PINN and GINN. The technical implementation framework of SIGD: the WANGZHEN SHENCE platform, is presented as a solution for the new pattern of complex system simulation, then the crowd simulation is used to explore the construction of "WANG ZHEN" system.

Key words: complex system, simulation intelligence, simulation intelligence based generating decisions (SIGD), PINN, surrogate model

中图分类号:

TP301.6

陈彬,郭润康,朱正秋等 . 一种复杂系统仿真研究的新模式SIGD[J]. 系统仿真学报, 2024, 36(12): 2993-3012.

Chen Bin,Guo Runkang,Zhu Zhengqiu,et al . A Novel Research Pattern for the Simulation of Complex Systems SIGD[J]. Journal of System Simulation, 2024, 36(12): 2993-3012.

图/表 21

图1

图2

图3

图4

图5

图6

图7

图8

图9

图10

图11

图12

图13

图14

图15

图16

图17

图18

表1

表2

图19

参考文献 31

1	王飞跃. 平行系统方法与复杂系统的管理和控制[J]. 控制与决策, 2004, 19(5): 485-489, 514.
	Wang Feiyue. Parallel System Methods for Management and Control of Complex Systems[J]. Control and Decision, 2004, 19(5): 485-489, 514.
2	Tuegel E J, Ingraffea A R, Eason T G, et al. Reengineering Aircraft Structural Life Prediction Using a Digital Twin[J]. International Journal of Aerospace Engineering, 2011, 2011(1): 154798.
3	Lavin Alexander, Krakauer David, Zenil Hector, et al. Simulation Intelligence: Towards a New Generation of Scientific Methods [EB/OL]. (2022-11-27) [2024-04-28]. .
4	戴汝为. 系统科学及系统复杂性研究[J]. 系统仿真学报, 2002, 14(11): 1411-1416.
	Dai Ruwei. Research on System Science and System Complexity[J]. Journal of System Simulation, 2002, 14(11): 1411-1416.
5	Lu Xiaoming, Wang Xiaoguang. Incorporating Heisenberg's Uncertainty Principle into Quantum Multiparameter Estimation[J]. Physical Review Letters, 2021, 126(12): 120503.
6	李劲, 肖人彬. 涌现计算综述[J]. 复杂系统与复杂性科学, 2015, 12(4): 1-13.
	Li Jin, Xiao Renbin. Emergent Computation: An Overview[J]. Complex Systems and Complexity Science, 2015, 12(4): 1-13.
7	纳西姆·尼古拉斯·塔勒布. 反脆弱: 从不确定性中获益[M]. 雨珂, 译. 北京: 中信出版社, 2020.
	Nassim Nicholas Taleb. Antifragile: Things That Gain from Disorder[M]. Translated by Yu Ke. Beijing: CITIC Publishing House, 2020.
8	王红卫, 李珏, 刘建国, 等. 人机融合复杂社会系统研究[J]. 中国管理科学, 2023, 31(7): 1-21.
	Wang Hongwei, Li Jue, Liu Jianguo, et al. Research on Human-machine Integration Complex Social System[J]. Chinese Journal of Management Science, 2023, 31(7): 1-21.
9	Feynman Richard P, Lebowitz Joel L. Statistical Mechanics: A Set of Lectures[J]. Physics Today, 1973, 26(10): 51-53.
10	Felix J W. The Theory of Symmetry Breaking in Continuous Structural Phase Transitions Hide affiliations[D]. The City of Provo: Brigham Young University, 1984.
11	康旭, 张晓峰. 基于生成对抗神经网络的雷达遥感数据增广方法[J]. 系统仿真学报, 2022, 34(4): 920-927.
	Kang Xu, Zhang Xiaofeng. Radar Remote Sensing Data Augmentation Method Based on Generative Adversarial Network[J]. Journal of System Simulation, 2022, 34(4): 920-927.
12	安小米, 马广惠, 宋刚. 综合集成方法研究的起源及其演进发展[J]. 系统工程, 2018, 36(10): 1-13.
	An Xiaomi, Ma Guanghui, Song Gang. Origins and Evolution of Meta-synthesis Approach[J]. Systems Engineering, 2018, 36(10): 1-13.
13	费敏锐, 狄轶娟. 仿真是第三种认识世界的基本活动[J]. 系统仿真学报, 2010, 22(2): 546-548.
	Fei Minrui, Di Yijuan. Third Basic Activity for Knowing World-simulation[J]. Journal of System Simulation, 2010, 22(2): 546-548.
14	Cybenko G. Approximation by Superpositions of a Sigmoidal Function[J]. Mathematics of Control Signals and Systems, 1989, 2(4): 303-314.
15	Penwarden M, Shandian Zhe, Narayan A, et al. Multifidelity Modeling for Physics-informed Neural Networks (PINNs)[J]. Journal of Computational Physics, 2022, 451: 110844.
16	Hall E J, Taverniers Søren, Katsoulakis M A, et al. GINNs: Graph-informed Neural Networks for Multiscale Physics[J]. Journal of Computational Physics, 2021, 433: 110192.
17	Pang Guofei, Lu Lo, Karniadakis G E. fPINNs: Fractional Physics-informed Neural Networks[J]. SIAM Journal on Scientific Computing, 2019, 41(4): A2603-A2626.
18	Cuomo Salvatore, Vincenzo Schiano Di Cola, Giampaolo Fabio, et al. Scientific Machine Learning Through Physics-informed Neural Networks: Where We are and What's Next[J]. Journal of Scientific Computing, 2022, 92(3): 88.
19	Aliakbari M, Mahmoudi M, Vadasz P, et al. Predicting High-fidelity Multiphysics Data from Low-fidelity Fluid Flow and Transport Solvers Using Physics-informed Neural Networks[J]. International Journal of Heat and Fluid Flow, 2022, 96: 109002.
20	刘霞, 冯文晖, 连峰, 等. 基于物理信息神经网络的气动数据融合方法[J]. 空气动力学学报, 2023, 41(8): 87-96.
	Liu Xia, Feng Wenhui, Lian Feng, et al. Aerodynamic Data Fusion Method Based on Physics-informed Neural Network[J]. Acta Aerodynamica Sinica, 2023, 41(8): 87-96.
21	孙昌璞. 量子力学诠释与波普尔哲学的"三个世界"[J]. 中国科学院院刊, 2021, 36(3): 296-307.
	Sun Changpu. Interpretation of Quantum Mechanics and Karl Popper's Three Worlds[J]. Bulletin of Chinese Academy of Sciences, 2021, 36(3): 296-307.
22	Bruel J M, Combemale B, Guerra E, et al. Comparing and Classifying Model Transformation Reuse Approaches Across Metamodels[J]. Software and Systems Modeling, 2020, 19(2): 441-465.
23	陈彬. 面向DEVS的多范式建模与仿真关键技术研究与实现[D]. 长沙: 国防科学技术大学, 2010.
	Chen Bin. The Research and Implementation of Key Techniques in DEVS Oriented Multi-paradigm Modeling and Simulation[D]. Changsha: National University of Defense Technology, 2010.
24	Vangheluwe H L M. DEVS as a Common Denominator for Multi-formalism Hybrid Systems Modelling[C]//CACSD. Conference Proceedings. IEEE International Symposium on Computer-Aided Control System Design. Piscataway: IEEE, 2000: 129-134.
25	Chen Bin, Zhong Ronghua, Vangheluwe Hans. Integrating Base Object Model Components into DEVS-based Simulation[J]. The Journal of Defense Modeling and Simulation Applications Methodology Technology, 2010, 7(4): 241-256.
26	Schulz S, Ewing T C, Rozenblit J W. Discrete Event System Specification (DEVS) and StateMate StateCharts Equivalence for Embedded Systems Modeling[C]//Proceedings Seventh IEEE International Conference and Workshop on the Engineering of Computer-Based Systems (ECBS 2000). Piscataway: IEEE, 2000: 308-316.
27	Carl Edward Rasmussen, K I Williams Christopher. Gaussian Processes for Machine Learning[M]. Cambridge: The MIT Press, 2005.
28	Yang Ling, Zhang Zhilong, Song Yang, et al. Diffusion Models: A Comprehensive Survey of Methods and Applications[J]. ACM Computing Surveys, 2023, 56(4): 105.
29	Jiang Yanqun, Chen Bokui, Li Xi, et al. Dynamic Navigation Field in the Social Force Model for Pedestrian Evacuation[J]. Applied Mathematical Modelling, 2020, 80: 815-826.
30	Helbing Dirk, Molnár Péter. Social Force Model for Pedestrian Dynamics[J]. Physical Review E, 1995, 51(5): 4282-4286.
31	Mohamed A, Qian Kun, Elhoseiny M, et al. Social-STGCNN: A Social Spatio-Temporal Graph Convolutional Neural Network for Human Trajectory Prediction[C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Piscataway: IEEE, 2020: 14412-14420.

数据集	ST-GCNN	ST-PINN-GCN	准确率提升/%
ETH	1.23	1.01	17.89
HOTEL	1.16	1.11	4.31
UNIV	0.55	0.52	5.45
ZARA1	0.63	0.52	17.46
ZARA2	0.60	0.55	8.33
AVG	0.83	0.74	10.84

数据集	ST-GCNN	ST-PINN-GCN	准确率提升/%
ETH	2.71	1.24	54.24
HOTEL	4.32	0.48	88.89
UNIV	1.46	0.55	62.33
ZARA1	2.15	0.92	57.21
ZARA2	1.53	0.84	45.10
AVG	2.43	0.81	66.67

[1]	田祥瑞, 尹婕, 姚睿, 万晓冬. 基于共生理论的复杂作战系统建模与分析[J]. 系统仿真学报, 2024, 36(1): 203-219.
[2]	刘文正, 张和明. 面向复杂系统运动仿真的信息物理融合建模[J]. 系统仿真学报, 2022, 34(5): 954-963.
[3]	胡晓峰, 齐大伟. 智能化兵棋系统：下一代需要改变的是什么[J]. 系统仿真学报, 2021, 33(9): 1997-2009.
[4]	朱峰, 姚益平, 唐文杰, 李进. 复杂系统仿真模型可重用性分析与应用[J]. 系统仿真学报, 2021, 33(10): 2356-2362.
[5]	邱晓刚, 陈亚洲, 张鹏. 从系统仿真到领域仿真的拓展[J]. 系统仿真学报, 2020, 32(9): 1637-1644.
[6]	胡春生, 许承东, 张鹏飞. 一种高度可配置的模块化仿真系统设计模式[J]. 系统仿真学报, 2020, 32(4): 627-637.
[7]	王帅, 朱峰, 姚益平, 唐文杰, 肖雨豪. 基于容器的复杂系统仿真应用自动封装技术[J]. 系统仿真学报, 2020, 32(11): 2199-2207.
[8]	鞠儒生, 蔡子民, 杨妹, 王松. 基于主成分分析的仿真结果评估方法[J]. 系统仿真学报, 2019, 31(12): 2678-2684.
[9]	张烙兵, 阮启明, 邱晓刚. 研究社会复杂系统的两种互补方法：仿真与博弈论[J]. 系统仿真学报, 2019, 31(10): 1960-1969.
[10]	姬杭, 周军华, 施国强, 林廷宇, 薛俊杰. 一种复杂系统轻量化建模与仿真技术框架[J]. 系统仿真学报, 2018, 30(6): 2013-2019.
[11]	耿超, 曲世友, 林廷宇, 施国强, 肖莹莹, 王玫, 马秋丹, 姬杭. 复杂数字化工业体系成熟度评估技术研究[J]. 系统仿真学报, 2018, 30(6): 2020-2026.