iCPS-DL: A Description Language for Autonomic Industrial Cyber-Physical Systems

TL;DR

The paper addresses automatic, safe reconfiguration of industrial CPS through iCPS-DL, a semantic description language that models physical/cyber-physical components, state estimation, and agent interactions. It introduces an ontology-driven autonomic supervisor that reasons over a state-estimation graph to select and deploy communicating agents, ensuring deadlock-freedom and liveness via behavioural types. A formal ontology meta-schema and a Water Distribution Network domain demonstrate how domains are defined, how estimators and sensors are coordinated, and how reconfiguration adapts to events such as sensor failures. The work provides a pathway to autonomous CPS management with formal reasoning, an open-source PoC, and a future-proof toolchain for code generation, deployment, and broader domain coverage.

Abstract

Modern industrial systems require frequent updates to their cyber and physical infrastructures, often demanding considerable reconfiguration effort. This paper introduces the industrial Cyber-Physical Systems Description Language, iCPS-DL, which enables autonomic reconfigurations for industrial Cyber-Physical Systems. The iCPS-DL maps an industrial process using semantics for physical and cyber-physical components, a state estimation model, and agent interactions. A novel aspect is using communication semantics to ensure live interaction among distributed agents. Reasoning on the semantic description facilitates the configuration of the industrial process control loop. A Water Distribution Networks domain case study demonstrates iCPS-DL's application.

Figures (9)

• Figure 1: The architecture of the autonomic reconfiguration framework. Industrial process, $P$, has a state vector $X$ and inputs actuator signal vector $W$. A distributed network connects multiple heterogeneous hardware components that perform control, monitoring, and optimisation tasks. Hardware components are depicted with white dashed rectangles. Controller, estimators, and sensor agents are depicted as rhombus, square, and circle shapes, respectively.
• Figure 2: The drinking water distribution network consists of a pump, a tank, and a junction where water is consumed. Three sensors monitor the tank inflow, tank water level, and actual water demand, along with a controller for the pump actuator. Each sensor agent communicates and exchanges information with other CPS components.
• Figure 3: The iCPS-DL grammar for defining local protocols, local configurations, global protocols, global configurations (above). An iCPS-DL snippet semantically describes a local and a global configuration for a control loop involving a tank estimator (below).
• Figure 4: Class diagram of the Water Distribution Network domain together with iCPS-DL description of the agent repository.
• Figure 5: Top: Graphical representation of the running example in Section \ref{['subsec:running_example']}. Middle: Graphical representation of the state estimation graph. Bottom: Graphical representation of the seven estimation trees rooted at ${\mathsf{tank}}\xspace.{\mathsf{head}}\xspace$, with a hardware device assignment for each node.

Theorems & Definitions (13)

• Example A.1: Sensor and estimator agents
• Example A.2: Loop transitions
• Example A.3: A local configuration for the ${\mathsf{WDN_{{\color{mymauve} {\mathsf{sim}}\xspace}}}}$
• Example A.4
• Theorem A.1
• proof
• Definition A.1: Projection/Composition relation
• Theorem A.2
• proof
• Theorem A.3