Integrating complex, heterogeneous large-scale systems.Future CPS willcontain heterogeneous distributed components and systems of large numbers that mustwork together effectively to deliver expected performance. There are several challenges toachieving this today. A fundamental issue is the lack of common terminology, modelinglanguages, and rigorous semantics for describing interactions—physical andcomputational—across heterogeneous systems. Achieving the interoperability andcompositionality of various components constructed in different engineering domains andsectors, without the benefit of unifying theories and standards, presents a major challenge.A lack of clear ownership of the interface between systems (e.g., between code, hardware,and multiple equipment vendors) also contributes to interoperability and integrationproblems. in addition to standards, interoperable systems need to ensure that timelyoutputs, outcome agreements, resilience, data transfers, and technical security protocolsare addressed seamlessly within and between components. This includes aggregating andsharing data within systems as well as across systems and components.Interaction between humans and systems.Current models for human andmachine behaviors are not adequate for designing CPS when humans and machinesclosely interact. One of the challenges is modeling and measuring situational awareness—human perception of the system and its environment and changes in parameters that arecritical to decision-making. This is particularly necessary for complex, dynamic systems,such as those used in aviation, air traffic control, power plant operations, militarycommand and control, and emergency services. in such systems situational awareness caninvolve large and unpredictable combinations of human and machine behavior. inadequatesituational awareness and limited ability to model the human component in large complexsystems has been identified as one of the primary factors in accidents related to humanerror (Nullmeyer et al, 2005).Dealing with uncertainty.Complex CPS need to be able to evolve and operatereliably in new and uncertain environments. An increasing number of these systems willalso demonstrate emergent and unknown behaviors as they become more and more relianton machine learning methodologies. in both cases, uncertainty in the knowledge oroutcome of a process will require new ways to quantify uncertainty during the CPS designand development stages. Current methods for characterization and quantification ofuncertainty are limited and inadequate. This is exacerbated by the limits of reliability andaccuracy of physical components, the validity of models characterizing them, networkconnections, and potential design errors in software. Ongoing debate also surrounds theexpectations for quantifying uncertainty, that is, attaining perfect results given theuncertainty of the physical world and approximations in design.