FORCES Overview - Shankar Sastry
Abstract:
Cyber-Physical Systems (CPS) are being increasingly deployed in critical infrastructures such as electric- power, water, transportation, and other networks. These deployments are facilitating real-time monitoring and control by exploiting the advances in wireless sensor-actuator networks, the internet of “everything”, data-driven analytics, and machine-to-machine interfaces. CPS operations depend on the synergy of com- putational and physical components. In addition, in many cases, CPS also interact with human decision makers. Fundamentally, once we admit that CPS operations depend on actions of humans (albeit to differ- ent degrees), we also have to admit that malicious entities could take charge of CPS control, by exploiting cyber insecurities or physical faults, or their combination. Therefore, to improve CPS resilience, we need diagnostic tools and control algorithms that ensure survivability in the presence of both, security attacks and random faults, and include the models of incentives of human decision makers in the design process. Resilient operation of CPS requires the following high confidence attributes: functional correctness (by design) for real-time operations, robustness to reliability failures (fault tolerance), and survivability even during successful attacks (operation through attacks). Designers and operators of CPS currently lack comprehensive tools for resilient operation. Major challenges include: (i) spatio-temporal and hybrid dynamics of cyber-physical processes; (ii) large number of interactions with interdependencies; and (iii) effects of public and private uncertainties. Two distinct domains of tools have emerged to respond to these challenges:
- Robust Control over networks: These tools primarily address safety and performance issues in closed- loop control over sensor-actuator networks.
- Theory of Incentives: These tools provide means to analyze and influence the strategic interactions of human decision makers.
To date, control and incentive tools have been designed and implemented separately, but modern CPS no longer permit this separation. Control and incentive tools designed in isolation, or without cognizance of strategic interactions between private entities and interdependent processes in CPS, are inadequate to maintain resilience. FORCES is developing an integrated resilient design methodology by adopting a rigorous analytical approach to allow the co-design of control and incentive tools. This will enable designers and operators to build-in resilience in CPS by maintaining synergistic integrations of human-centric elements with automated diagnostic and control processes.