不确定环境下信息物理系统高效可信构造关键技术研究

As a key technology of Intelligent Manufacturing, Cyber-Physical Systems (CPSs) are supposed to be able to accurately and efficiently conduct the sensing, actuation, communication and control within physical world. Since CPS designs involve both the uncertain physical environments and parallel executing devices with performance variations which interact with each other intensely, the design complexity of CPSs is extremely high. If a CPS product is not sufficiently verified and optimized in the design phase, it is hard for us to assure that the product can execute correctly and efficiently while satisfying the design requirement in practice. In other words, the trustworthiness of the CPS design cannot be guaranteed within an uncertain environment. ..This project tries to propose a comprehensive top-down design methodology for trustworthy CPS products from the perspectives of three major construction steps, i.e., modeling/specification, design and implementation. The objective of this project is to explore efficient and effective modeling, evaluation, synthesis, simulation and verification techniques for CPS designs considering uncertain physical environment. Aiming at reducing the overall CPS construction time as well as improving the quality of the design, this project will mainly investigate the following three issues: 1) How to enable automated evaluation and synthesis of CPS models/specifications within uncertain environment? This project will establish a contract-based modeling and analysis framework for CPSs, which enables automated evaluation and synthesis of uncertain CPS designs. Therefore, the CPS design quality as well as CPS development process can be optimized. 2) How to improve the simulation performance of CPS virtual prototypes? This project will utilize various multi-granularity parallel simulation techniques to reduce the performance evaluation and design space exploration time for virtual prototypes. 3) How to guarantee the refinement correctness of different abstractions of CPS designs? This project will study how to reuse the validation efforts among design models/specifications, virtual prototypes, and low-level implementations of CPS designs, which can not only enable the functional consistency checking among these abstractions, but also reduce the overall validation efforts. ..In conclusion, by exploring various efficient and effective evaluation, synthesis, simulation and verification techniques, this project will provide a promising methodology and tool framework to support the efficient construction of trustworthy CPSs. The outcomes of this project will not only significantly reduce the CPS construction time, but also enhance the performance, reliability and predictability of constructed CPSs.

作为智能制造的核心技术，信息物理系统（CPS）能够对物理世界精确地实施“感、执、传、控”。然而由于所处环境与系统器件存在多种不确定性，CPS构造过程复杂度极高，设计缺乏充分的验证极易导致最终产品的可信难以保证。本项目围绕CPS可信构造，从规约、设计、实现三个层次出发，探索不确定环境下CPS系统的高效建模、评估、综合、仿真与验证等关键构造技术，主要研究内容包括：1）研究不确定环境下基于契约的CPS设计理论及其自动评估与综合方法，提高CPS设计的质量与开发效率；2）研究不确定环境下虚拟原型的多粒度并行仿真方法，降低不确定环境感知虚拟原型的性能评测时间；3）研究CPS设计规约、虚拟原型与底层实现间验证结果重用机制，确保各抽象层次间功能的一致性，并降低系统的总体验证代价。本项目的研究将为不确定环境下CPS的可信构造提供共性理论与技术支撑，在提高CPS质量与可信的同时，大幅降低其构造所需的时间。

虽然信息物理系统（CPS）能够对物理世界精确地实施“感、执、传、控”，但由于所处环境与系统器件存在多种不确定性，其构造过程复杂度极高，设计缺乏充分的验证极易导致最终产品的可信难以保证。本项目围绕CPS可信构造，从规约、设计、实现三个层次出发，研究了不确定环境下CPS系统的高效建模、评估、优化、综合、仿真与验证等关键构造技术，主要创新如下:..1）不确定环境下CPS的建模与量化评估：扩展UML顺序图语法语义，支持各类不确定CPS场景的建模。提出覆盖顺序图所有语法结构的转化模板，支持不确定UML顺序图至网络价格时间自动机的自动转化，支持基于统计模型检验的定量评估。.2）CPS系统级规约的自动化综合与并行验证：基于部分给定实例以及不完全的时钟约束规约（CCSL）模板，通过将其分别转化为SKETCH问题与基于强化学习的填空枚举问题，实现了多种用户实例制导的CCSL规约快速自动化综合。针对CPS受限连续系统特性，提出了基于两阶段求解的障碍证书（Barrier Certificate，BC）快速求解方法，通过并行协同求解的方式将传统BC求解时间降低了几个数量级。.3）CPS规约与实现的一致性检测：为确保虚拟原型与物理设备功能的一致性，提出了基于变异的一致性检测框架。基于新提出的变异算子，通过符号化执行支持自动生成有效测试用例，支持高层系统级规约和低层级具体实现之间的一致性检测。.4）不确定环境下CPS的设计优化：针对由软错误导致的CPS部件性能差异，基于动态电压调整（DVFS）机制提出了多种面向云端与设备端的电压分配优化方案，在保证CPS时间可靠性的基础上降低系统能耗。针对异构设备数据分布的不确定性，提出高效联邦学习算法框架，实现从分类精度、实时响应、鲁棒性等多个角度的优化。针对交通CPS动态运行不确定场景，提出了多种基于强化学习的算法，提升交通信号控制效果。.. 本项目的研究为不确定环境下CPS的可信构造提供共性理论与技术支撑，在提高CPS功能与性能可信的同时，大幅降低其总体构造的时间。相关方法与技术已在智慧交通等领域开展应用验证。