系统控制之美论坛

夏元清会士：中原工学院、北京理工大学

       简介：夏元清，男，54岁，中共党员，控制科学与工程领域专家。国家杰出青年基金获得者，教育部长江学者特聘教授，国家“万人计划”领军人才，IEEE Fellow，CAA Fellow，享受国务院特殊津贴专家。现任中原工学院校长，北京理工大学讲席教授；国务院学位委员会第八届学科评议组成员，中国计算机学会大数据专家委员会委员，中国仪器仪表学会物联网工作委员会副理事长，中国指挥与控制学会云控制与决策专业委员会主任委员，天基智能信息处理全国重点实验室副主任。长期致力于多源信息复杂系统的信息处理与控制、飞行器控制、无人移动平台协同控制、空天地海一体化网络环境下多运动体系统跨越协同控制与智能决策、云控制与决策等研究，承担国家重点研发计划、国家自然科学基金重点项目、973计划等多层次科研项目多项。在国际上首次提出“云控制”概念，并建立相关理论与技术体系，赋能天空地海有人、无人系统，智能制造等领域，极大提升了感知、决策与控制品质。获国家科技进步二等奖1项，教育部自然科学二等奖2项，北京市科学技术二等奖3项，国防科技进步三等奖1项，吴文俊人工智能自然科学奖一等奖1项、二等奖1项，中国自动化学会自然科学一等奖1项、科技进步一等奖1项，中国指挥与控制学会科学技术奖一等奖1项等。发表学术论文1000余篇，其中SCI收录800余篇，申请发明专利100余件，出版英文专著22部，中文专著3部，论文累计被引38000余次；2014年至今连续入选Elsevier中国高被引学者榜单；入选斯坦福全球前2%顶尖科学家终身科学影响力榜单。荣获“2024年河南省最美科技工作者”等荣誉称号。

题目：智能网联无人系统云控制理论与技术

Title：Cloud Control Theory and Technology for Intelligent and Networked Unmanned Systems

        摘要：智能网联无人系统云控制是未来群体智能机器人产业发展的战略方向。由于群体智能机器人运行的复杂性和特殊性，上下游企业之间、不同行业之间，以及智能机器人各体间都存在信息壁垒，严重限制了高级别群体智能机器人的发展。全域感知和精准控制作为智能网联无人系统云控制的关键核心问题，目前存在数据融合与数字孪生的瓶颈。随智能机器人规模增加，个体之间强耦合特性与多约束特性更加凸显，严重影响算法计算效率和实际性能。基于误差反馈的控制方法难以适应动态拓扑变化的云控制问题，且面临大规模、强耦合、多约束、解算慢等挑战，因此亟需研究智能网联无人系统云控制方法，实现复杂环境下群体智能机器人实时动态精准控制。智能网联无人系统云控制理论与技术研究聚焦人工智能、智能网联与低空经济等战略性新兴领域，紧密围绕国家“发展新质生产力”和“加快现代化产业体系建设”的政策导向，积极服务中国式现代化与区域高质量发展。推动智能网联无人系统云控制的落地进程，对我国抢占世界科技创新和产业迭代高地具有重要意义。

     Abstract: Cloud control for intelligent connected unmanned systems is a strategic direction for the future development of the swarm intelligent robot industry. Due to the complexity and particularity of swarm intelligent robot operations, there are information barriers between upstream and downstream enterprises, between different industries, and among individual intelligent robots, which severely restrict the development of high-level swarm intelligent robots. As the key core issues of cloud control for intelligent connected unmanned systems, global perception and precise control currently face bottlenecks in data fusion and digital twin. With the expansion of the scale of intelligent robots, the strong coupling and multi-constraint characteristics among individuals become more prominent, which seriously affects the algorithm's computing efficiency and actual performance. The control method based on error feedback is difficult to adapt to the cloud control problem with dynamic topology changes, and faces challenges such as large scale, strong coupling, multiple constraints, and slow solution. Therefore, it is urgent to study the cloud control method for intelligent connected unmanned systems to realize real-time, dynamic and precise control of swarm intelligent robots in complex environments. The research on the theory and technology of cloud control for intelligent connected unmanned systems focuses on strategic emerging fields such as artificial intelligence, intelligent networking and low-altitude economy. It closely aligns with the national policy orientations of "developing new quality productive forces" and "accelerating the construction of a modern industrial system", and actively serves Chinese-style modernization and high-quality regional development. Promoting the implementation of cloud control for intelligent connected unmanned systems is of great significance for China to seize the high ground of global scientific and technological innovation and industrial iteration.

张艳霞：北京理工大学  

题目：关于线性子空间之间的角度度量

Title：on the angular metrics between linear subspaces

王琳：上海交通大学

        简介：王琳，上海交通大学教授、博导，上海市曙光学者，围绕网络化系统的分析与控制、大规模集群系统的调度优化开展工作，发表SCI期刊论文60余篇，获2022年中国自动化学会自然科学奖一等奖，2022年上海市科学技术奖自然科学二等奖，2020年中国电子学会电子信息领域优秀科技论文奖，第13届智能控制与自动化世界大会最佳理论论文奖。担任IFAC大规模复杂系统专委会副主席、中国工业与应用数学学会复杂网络与复杂系统专委会副主任，Systems & Control Letters副编委, J. Systems Science & Complexity青年编委，《系统科学与数学》和《指挥与控制学报》编委。

题目：自动驾驶测试场景生成与综合评测方法研究

Title：Study on Test Scenario Generation and Comprehensive Evaluation Methods for Autonomous Driving

        摘要：科学的测试与评价是推动自动驾驶技术进步的重要基础和核心保障。随着大量人工智能算法融入自动驾驶的感知、决策等模块，对测试评价方法的精度和效率等方面提出了新需求。由于关键测试场景的稀缺性、复杂性和高维性，如何有效生成用于自动驾驶测试的关键测试场景已成为一个重大挑战。本报告旨在为关键边界测试场景提供一个具有灵活复杂度和多样性的在线自适应生成框架，以测试自动驾驶汽车的综合性能。在此基础上，探讨用于自动驾驶车辆综合性能评估的客观多维度综合评测方法。为了保证自动驾驶车辆评测的客观性和测试效率，提出了一套完整的包含两个评测维度、五个一级指标、十四个二级指标的综合评价体系。为了实现全流程自动化的评测，提出了基于场景复杂度模型的客观评测指标权重计算方法和可缩放的评测结果计算方法。所开发的自动化评测体系和评测方法可以同时应用于自动驾驶车辆仿真测试、硬件在环测试、实车在环测试以及实车测试，可以有效提升测试效率，加速智能驾驶车辆的实际应用部署。

      Abstract: Scientific testing and evaluation are important foundations and core guarantees for promoting the advancement of autonomous driving technology. With the integration of a large number of artificial intelligence algorithms into the perception, decision-making and other modules of autonomous driving, new requirements have been put forward for the accuracy and efficiency of testing and evaluation methods. Due to the scarcity, complexity and high dimensionality of key test scenarios, how to effectively generate key test scenarios for autonomous driving testing has become a major challenge. This report aims to provide an online adaptive generation framework with flexible complexity and diversity for key boundary test scenarios to test the comprehensive performance of autonomous vehicles. On this basis, it explores an objective multi-dimensional comprehensive evaluation method for the comprehensive performance assessment of autonomous vehicles. To ensure the objectivity and testing efficiency of autonomous vehicle evaluation, a complete comprehensive evaluation system is proposed, which includes two evaluation dimensions, five first-level indicators and fourteen second-level indicators. To realize the full-process automated evaluation, an objective evaluation index weight calculation method based on the scenario complexity model and a scalable evaluation result calculation method are proposed. The developed automated evaluation system and evaluation methods can be applied to autonomous vehicle simulation testing, hardware-in-the-loop testing, vehicle-in-the-loop testing and real vehicle testing simultaneously, which can effectively improve testing efficiency and accelerate the practical application and deployment of intelligent driving vehicles.

穆义芬：中国科学院数学与系统科学研究院 

      简介： 穆义芬，中国科学院数学与系统科学研究院副研究员，博士生导师。研究方向为博弈论与博弈学习理论。2005年于北京大学数学科学学院获理学学士学位、哲学双学士学位；2010年于中科院数学与系统科学研究院获理学博士学位。现为中国运筹学会博弈论分会理事、CCF计算经济学专委会执委、中科院数学与系统科学研究院“博弈学习演化与自适应控制机制”团队负责人、国家数学与交叉科学中心数学与信息技术交叉研究部“复杂系统博弈理论”方向联合负责人、期刊 Journal of Dynamics and Games编委、期刊Journal of Advanced Computational Intelligence and Intelligent

Informatics (JACIII） Associate Editors (AE). 曾获“关肇直青年研究奖”，研究论文获2024年关肇直奖（2/1672），2023年the 8th International Workshop on Advanced Computational Intelligence and Intelligent Informatics(IWACIII) Young Researcher Award，2022 年IEEE CSS Beijing Chapter Young Author Nomination Prize等。

题目：对学习算法的最优对策及重复博弈中的异质动力学：从人机博弈开始

Title：The algorithm exploitation and heterogeneous learning dynamics: all from human-machine games

        摘要：随着互联网、大数据、算法与人工智能等技术不断发展，社会进入智能系统时代，大量算法辅助或参与决策特别是博弈决策。当算法作为主体进行博弈时，人机博弈成为一个自然的研究课题，对其研究具有多方面意义，既可以帮助改进设计算法，又可以为算法规制提供理论支持。大约20年前，基于控制论与博弈论相结合的思想，郭雷院士极其富有前瞻性地首倡研究人机博弈与博弈控制系统。在这一框架下，数学院博弈论研究团队得到了一系列具有特色的研究成果，如对经典机器策略或学习算法的最优对策、最优对策下系统的周期与拟周期行为、基于人机博弈框架的异质博弈动力学以及相应的纳什均衡新解法等。目前这一研究课题受到计算机、控制论、博弈论等许多领域学者越来越多的关注，研究也不断拓展和深入。本报告将介绍数学院师生在郭雷院士指导下取得的相关研究成果，并对未来研究尝试展望！

      Abstract: With the continuous development of technologies such as the Internet, big data, algorithms, and artificial intelligence, society has entered the era of intelligent systems. A large number of algorithms assist or participate in decision-making, especially game decision-making. When algorithms act as agents to conduct games, human-machine game has become a natural research topic. Its research is of great significance in multiple aspects: it can not only help improve and design algorithms, but also provide theoretical support for algorithm regulation. Approximately 20 years ago, based on the idea of combining cybernetics and game theory, Academician Guo Lei took a highly forward-looking initiative to advocate the research on human-machine game and game control systems. Under this framework, the game theory research team of the School of Mathematics has obtained a series of distinctive research results, such as the optimal countermeasures against classical machine strategies or learning algorithms, the periodic and quasi-periodic behaviors of the system under optimal countermeasures, the heterogeneous game dynamics based on the human-machine game framework, and the corresponding new solution methods for Nash equilibrium. Currently, this research topic has attracted increasing attention from scholars in many fields such as computer science, cybernetics, and game theory, and the research is continuously expanding and deepening. This report will introduce the relevant research achievements made by teachers and students of the School of Mathematics under the guidance of Academician Guo Lei, and attempt to look forward to future research!

潘豪：中国航天科技集团一院十二所 

        简介：潘豪，男，高级工程师，主要从事运载火箭姿态控制系统设计与研究工作，现为型号主任设计师，先后参与完成多发运载火箭设计工作。

题目：运载火箭高可靠姿态控制技术

Title：High-Reliability Attitude Control Technology for Launch Vehicles

        摘要：基于从事的运载火箭研制工作，从提升运载火箭飞行可靠性的角度出发，介绍运载火箭大风区减载控制，弹性自适应增广控制，以及姿控喷管故障辨识与重构控制技术方法，并从试验验证分析角度，给出了方法的可行性和有效性。

      Abstract: Based on the research and development work of launch vehicles, from the perspective of improving the flight reliability of launch vehicles, this paper introduces the technical methods of launch vehicle's wind zone load reduction control, elastic adaptive augmented control, and attitude control nozzle fault identification and reconstruction control. From the perspective of test verification and analysis, the feasibility and effectiveness of the methods are presented.

丛新蓉：北京工商大学  

        简介：丛薪蓉，北京工商大学数学与统计学院应用统计系副教授，硕士生导师。博士毕业于哈尔滨工业大学数学系，随后于中国科学院数学与系统科学研究院完成博士后研究。长期致力于随机与不确定性动力系统的估计与控制，随机系统数值算法与仿真、风险决策与控制的研究。对于来自经济、金融、航空航天等领域的随机动力系统，利用随机微分方程、随机分析、数理统计等理论，对系统的可控性及稳定性进行研究。主持国家自然科学基金青年项目 1 项，参与国家自然科学基金面上项目3项。

题目：基于参数不确定性的机翼可压缩流动控制研究

Title：Study on Compressible Flow Control of Wings Based on Parametric Uncertainty

        摘要：跨音速机翼流动中存在的激波会导致阻力增加、升力降低乃至抖振等不利现象，主动流动控制是抑制这些现象的有效手段。然而，实际飞行环境中存在的参数不确定性，如来流马赫数、攻角及气动外形等的随机波动，会显著影响控制的鲁棒性与可靠性。为此，针对某典型机翼，开展考虑参数不确定性的可压缩流动控制研究。首先，通过求解雷诺平均Navier-Stokes方程，构建高保真度流场模型；在此基础上，引入主动控制策略作为控制手段。进而，采用多项式混沌展开等不确定性量化方法，系统量化关键不确定参数对系统输出的影响程度与统计规律。最终，基于量化分析结果，设计并评估了一种鲁棒控制策略。研究结果表明，相较于传统的确定性控制设计，所提出的鲁棒控制方法在参数存在摄动时，能更稳定地维持控制效果，显著提升了机翼在不确定飞行环境下的气动性能与安全性能。

      Abstract: Shock waves in transonic wing flow can lead to adverse phenomena such as increased drag, reduced lift, and even buffeting. Active flow control is an effective means to suppress these phenomena. However, parameter uncertainties in the actual flight environment, such as random fluctuations in freestream Mach number, angle of attack, and aerodynamic shape, will significantly affect the robustness and reliability of control. Therefore, for a typical wing, this study carries out research on compressible flow control considering parameter uncertainties. First, a high-fidelity flow field model is constructed by solving the Reynolds-averaged Navier-Stokes equations. On this basis, an active control strategy is introduced as the control method. Furthermore, uncertainty quantification methods such as polynomial chaos expansion are used to systematically quantify the influence degree and statistical laws of key uncertain parameters on the system output. Finally, based on the quantitative analysis results, a robust control strategy is designed and evaluated. The research results show that compared with the traditional deterministic control design, the proposed robust control method can maintain the control effect more stably when parameters are perturbed, and significantly improve the aerodynamic performance and safety performance of the wing in an uncertain flight environment.

陈辉：北京理工大学  

        简介：陈辉，2018年获山东大学数学学士学位，2023年获中国科学院数学与系统科学研究院系统理论博士学位。目前为北京理工大学博士后，主要致力于自适应模型预测控制、兵棋推演与智能决策等方向的交叉研究。摘要：模型预测控制（MPC）近年来在工业界和学术界获得了日益广泛的关注。传统的MPC设计依赖于特定的系统模型，其控制效果直接受模型精确度的影响。为应对系统存在的不确定性，自适应及基于学习的方法被引入到MPC设计中。本报告针对若干典型系统，设计了自适应MPC算法。在一定的假设条件下，通过结合加权最小二乘算法与随机正则化方法，我们不仅证明了自适应MPC的可行性，还验证了闭环控制性能的收敛性。然而，这些假设条件具有一定的局限性，针对这一问题，我们也尝试提出了若干切实可行的解决思路。

题目：自适应模型预测控制的可行性与收敛性

Title：Feasibility and Convergence of Adaptive Model Predictive Control

        摘要：模型预测控制（MPC）近年来在工业界和学术界均受到越来越多的关注。传统模型预测控制的设计依赖于特定的系统模型，其控制效果受模型精度影响。为应对系统不确定性，在模型预测控制设计中会采用自适应或基于学习的方法。本报告针对几类典型系统设计了自适应模型预测控制算法。在若干假设条件下，通过加权最小二乘算法和随机正则化方法，我们证明了自适应模型预测控制的可行性以及闭环控制性能的收敛性。但这些假设存在一定局限性，为此我们也尝试提出了几种可行的解决方法。

        abstract：Model predictive control (MPC) has recently attracted increasing attention in both industry and academia. The traditional MPC design relies on the specific system models, and its effectiveness is influenced by the accuracy of the models. To address system uncertainties, adaptive or learning-based approaches are employed in the design of MPC. In this report, adaptive MPC algorithms are designed for some typical systems. Under some assumptions, we establish both the feasibility of the adaptive MPC and the convergence of the closed-loop control performance by using the weighted least-square algorithm and the random regularization method. However, these assumptions have  certain limitations. To address these issues, we also attempted to provide several feasible approaches.

陈鸽，中国科学院数学与系统科学研究院研究员

报告题目：网络演化博弈研究进展

        陈鸽，中国科学院数学与系统科学研究院研究员。主要研究兴趣为多智能体自组织涌现理论。发表SIAM Review, IEEE TAC, Automatica，IEEE TMC等期刊论文多篇，承担基金委、科技部、中科院项目多项。曾获中国控制论会议“关肇直奖”，中国自动化学会自然科学奖一等奖，中国运筹学会“中国运筹学应用奖”一等奖，美国工业与应用数学学会SIGEST论文奖励，中科院数学与系统科学研究院“年度科研进展奖”等。

       摘要：网络演化博弈主要研究个体的博弈决策在网络上的动力学行为，包括群体行为的形成与演化机制。我们研究了二维网络上的囚徒困境、雪堆博弈（鹰鸽博弈，胆小鬼博弈）、和猎鹿博弈等模仿动力学模型，给出系统收敛的参数条件。此外，发现了收敛时间可能存在相变现象，并且在收敛到稳态之前会长时间停留在某个亚稳态。进一步，研究了新能源电力系统Stackelberg博弈问题，并给出了纳什均衡的一个解析解。