Automated System Testing and Repair for Cyber Physical Systems

自动驾驶汽车和物联网（IoT）可以在许多方面改善我们的生活。自动驾驶汽车有望消除90%的交通事故。物联网是关于通过互联网连接传感器，执行器和许多以前未连接的东西。它是许多重要应用的关键推动者，如应急和灾害管理系统、智能城市和联网车辆。尽管最近在自主和互联系统方面取得了进展，但由于这些系统的严重故障，例如自动驾驶汽车事故或救灾系统中的连接和性能问题，社会尚未完全信任这些系统。自主和互联系统具有复杂的网络物理特征。与其他网络物理系统（CPS）类似，为了确保其正确性，可靠性和鲁棒性，这些系统在整个开发生命周期中都要进行广泛的验证和测试。在过去的几年里，用于软件程序的验证、测试和错误修复的自动化技术取得了显著的进步。然而，对于今天的连接和自主CPS，需要系统级测试和错误修复技术，这些技术可以大规模应用，并可以处理这些系统的异构性，网络物理和连接方面。 该研究项目开发了CPS的自动化系统测试和修复方法，有效地利用了系统规范和编码到系统组件及其接口中的知识。该研究将建立在CPS开发过程中通常创建的现实输入和工件的基础上，并解决CPS领域的普遍情况，从而在可执行和特定于领域的建模语言（如Simulink）中指定系统行为设计，并通过专用的高保真模拟器进行早期测试。本研究的目标将通过新的，可扩展的和实用的技术，揭示，调试和修复CPS模型中的需求违规。这些技术在实现CPS的早期和系统测试和修复的同时，将解决CPS验证中的两个主要挑战：（1）处理CPS的复杂性和异质性，以及（2）揭示和修复与系统组件之间的不期望的交互有关的违规。 这项研究的成果将是CPS的工具支持和自动化系统级测试和修复方法，导致软件验证技术，可以在降低开发成本，增加可信度和提高创新方面发挥重要作用。我预计，在这项研究中开发的专业知识和工具将为与加拿大汽车和电信行业的长期，高影响力的合作铺平道路。该研究计划将进一步帮助培养许多学生和研究人员，使他们具备必要的知识和技能，以推动物联网和自主系统等新兴领域的研究和创新。

Autonomous cars and Internet-of-Things (IoT) can improve our lives in many ways. Self-driving cars are expected to eliminate 90% of accidents. IoT is about connecting sensors, actuators and many previously unconnected things through the Internet. It is the key enabler of many important applications such as emergency and disaster management systems, smart cities and connected vehicles. Despite recent advances in autonomous and connected systems, society does not yet fully trust these systems due to their critical failures such as examples of self-driving cars' accidents or connectivity and performance issues in disaster relief systems. Autonomous and connected systems have complex cyber physical traits. Similar to other cyber physical systems (CPS), to ensure their correctness, reliability and robustness, these systems are subject to extensive verification and testing throughout their entire development lifecycle. In the past years, automated techniques for verification, testing and bug fixing of software programs have progressed significantly. However, for today's connected and autonomous CPS, there is a need to system-level testing and bug fixing techniques that can be applied at scale and can handle heterogeneity, cyber physical and connectivity aspects of these systems. This research program develops automated system testing and repair methods for CPS, effectively leveraging system specifications and the knowledge encoded into system components and their interfaces. The research will build on realistic inputs and artifacts that are commonly created during CPS development and addresses a prevalent situation in the CPS domain, whereby the system behavioural design is specified in executable and domain-specific modeling languages such as Simulink and tested early on via dedicated, high-fidelity simulators. The goals of this research will be achieved through novel, scalable and practical techniques that reveal, debug and repair requirements violations in CPS models. These techniques, while enabling early and system testing and repair of CPS, will address two major challenges in CPS verification: (1) handling complexity and heterogeneity of CPS, and (2) revealing and repairing violations pertaining to undesired interactions among system components. The outcome of this research will be tool-supported and automated system-level testing and repair approaches for CPS, leading to software verification techniques that can have a major role in reducing development costs, increasing trustworthiness, and improving innovation. I anticipate that the expertise and tools developed in this research will pave the way to long-lasting, high-impact collaborations with the automotive and telecommunication sectors in Canada. The research program will further help train many students and researchers with the knowledge and skills necessary to push the boundaries of research and innovation in emerging fields such as IoT-enabled and autonomous systems.