Early Modelling of Interactions Between Humans, AI and Low Level Automation Towards System Resilience

The context of this research is the early design of Cyber Physical Systems of Systems (CPSoS) that include AI components for decision support and for high level system operation. Industry 5.0 recognises the increased importance of AI in CPSoSs and system resilience is required. The challenge is the high level design of a CPSoS so that it can handle safety or security related events. There is a need to model the interactions between the humans, the AI and the low level automation of CPSoS for system resilience (ability of the system to withstand disturbances and keep its operation or drive to a safe state). The proposed concept is a model driven approach based on dependency, state machine and sequence models to capture the topology, behavior and timing of system components. The flow of decisions starts from the human operators (possibly supported by AI decision support components), then to AI components that handle high level control actions and finally to low level automation controls. Disturbances can happen in any of these levels or from the external environment and the system needs to be designed to mitigate the effects and sustain its operation. UML use case diagrams demonstrate the use cases of the system while class diagrams describe the functional decomposition and the overall system structure, focusing on the dependencies between system elements (humans, AI, automation and interfaces to the environment). State machine diagrams capture the system behavior and sequence diagrams are used to model interactions. A methodology is proposed to identify how the system responds to reliability or threat events that affect different parts of the CPSoS