A framework for Model-Driven Engineering of resilient software-controlled systems

Abstract

Emergent paradigms of Industry 4.0 and Industrial Internet of Things expect cyberphysical systems to reliably provide services overcoming disruptions in operative conditions and adapting to changes in architectural and functional requirements. In this paper, we describe a hardware/software framework supporting operation and maintenance of software-controlled systems enhancing resilience by promoting a Model-Driven Engineering (MDE) process to automatically derive structural configurations and failure models from reliability artifacts. Specifically, a reflective architecture developed around digital twins enables representation and control of system Configuration Items properly derived from SysML Block Definition Diagrams, providing support for variation. Besides, a plurality of distributed analytic agents for qualitative evaluation over executable failure models empowers the system with runtime self-assessment and dynamic adaptation capabilities. We describe the framework architecture outlining roles and responsibilities in a System of Systems perspective, providing salient design traits about digital twins and data analytic agents for failure propagation modeling and analysis. We discuss a prototype implementation following the MDE approach, highlighting self-recovery and self-adaptation properties on a real cyber-physical system for vehicle access control to Limited Traffic Zones.

Self-Adaptation Driven by SysML and Goal Models -- A Literature Review

“…Parri et al [85] Combined system configurations derived from SysML block definition diagrams (BDD) metadata with a failure model derived from fault tree via digital twins and data analysis agents.…”

“…7. Real configuration items from SysML BDD properties and diagnostic, predictive, and prescriptive analytics derived from fault tree are integrated by Parri et al [85] to discover alternative configurations when a runtime violation is detected. 8.…”

“…To support system reliability, flexibility, and runtime recoverability, Parri et al [85] proposed a software/hardware framework, called JARVIS, for developing CPSs and Systems of Systems. JARVIS adopts SysML BDDs and fault trees to discover configuration alternatives using digital twins and data analytic agents.…”

“…From the Goal/SysML integration methods, Ingram et al [45] used fault tolerance analysis and rules to deal with errors. Without considering goal models, Ribeiro et al [91], Parri et al [85], Bareiß et al [84], Souza et al [93], and Soyler and Sala-Diakanda [17] triggered their embedded strategies, configuration, or plans to respond to internal or environmental changes. Similarly, Morandini et al [82,83] represented the goal model in an agent structure while embedding the environmental and failure conditions, alternatives, and plans in agent beliefs and system design.…”

“…However, they were not used in their strategy selection process triggering the possible solutions depending on several conditions attached to the system operations as rules. Similarly, Parri et al [85] used Fault Tree analysis to detect and predict failures while the suitable configurations were associated to the tree via digital twins. Also, Qian et al [94] used case-based reasoning in all MAPE activities (see Figure 1) except in the planning process, the latter being supported by goal-based reasoning when it failed.…”