From Chance to Choice: Strategies to Attaining Resilience in Cyber-Physical Systems

Doctoral thesis, 2024

Background: Autonomy is a key attribute of cyber-physical systems engineered to achieve human-machine coexistence and collaboration toward human-centered goals. To be trusted, autonomous systems must operate resiliently, yet designing and verifying resilient behavior remains an open challenge. Resilient cyber-physical systems should avoid, withstand, recover from, and adapt to adversities arising from computational, network, or environmental disruptions. Wearable biosensors are a prime example of cyber-physical systems that must operate resiliently. Such a healthcare monitoring system could fail during a network outage or erroneous sensor data, endangering lives. A resilient healthcare monitoring system, with redundant paths and adaptive capacity, ensures continuous monitoring and timely alerts despite disruptions.

Objective:This thesis aims to equip developers and quality assurance teams with strategies for attaining resilience in cyber-physical systems, ensuring that resilience is engineered rather than attained by coincidence. Attaining resilience in cyber-physical systems entails justified adaptation to overcome unknown stimuli, ever-changing objectives, and deprecated components. Software as a tool for self-management is crucial for dealing with uncertainty. Achieving resilience is challenging since unexpected effects may emerge during execution, requiring runtime decision-making rather than design time.

Method: The strategies are rooted in publications in software engineering, self-managed and adaptive systems, robotics, and transportation. They encompass quantitative and qualitative research that follows a design science research methodology.

Results: The thesis introduces seven strategies for attaining resilience, including:
(i) best practices for runtime assessment,
(ii) tools to manage interactions among diverse and smart agents,
(iii) methods for uncertainty mitigation at the code level, runtime adaptation, and explanation of property violations, and
(iv) exemplars that serve as models to advance resilience research.
Our results demonstrate that resilience is achieved through systematic design and runtime decision-making, ensuring that systems consistently meet operational goals.

Conclusion:
This study advocates for resilience as a strategic goal, highlighting its importance as a foundational discipline within software engineering for cyber-physical systems. The findings benefit both researchers and practitioners, emphasizing resilience engineering as essential for the future of autonomous systems.