Periodic Reporting for period 1 - DeFacto (Design Automation for Smart Factories)
Reporting period: 2020-10-01 to 2022-09-30
Manufacturing is undergoing what many identified as "a fourth industrial revolution": production lines are becoming complex networked systems, equipped with teams of robots that manipulate products and take care of the logistics and shop-floor organization, sensors gathering large amounts of data, and computing platforms providing situational awareness and intelligent control. "Smart" factories will turn into complex, heterogeneous cyber-physical systems (CPSs), a transformation that will offer unprecedented opportunities. Reconfigurable production lines will be able to face the increasing demand for customized products and move from mass production to mass customization, data analytics will enable optimizing production costs, supply and logistics chains, and integrating different aspects of the value production chain, from the sales office to the production line. Performing predictive maintenance at scale and reducing production downtime will be possible. Indeed, exploiting the opportunities provided by such innovations will be crucial to increasing the added value and productivity of European manufacturing and addressing the challenges faced by the European economy and society.
These opportunities will come, however, with a series of engineering challenges. A smart manufacturing system must integrate a diverse set of components while offering strong guarantees in terms of functionality, reliability, safety, and cost. This heterogeneity in components and system requirements inevitably calls for models, specification formalisms, and design constraints of different nature to represent a design space that is difficult to extensively explore in a reasonable time, an issue that is often recognized as the "explosion in complexity" of today's industrial CPS design.
Mirroring the success of electronic design automation (EDA) in taming the complexity of microchip design in the '80s, system design automation is expected to play a crucial role in reducing the complexity of CPS design. The scientific goal of DeFacto is to advance the state of the art in system design automation by developing novel modeling paradigms, scalable algorithms, and tools to aid the design of smart manufacturing systems. DeFacto aims at automatically defining the architecture (i.e. the system components and their interconnections) and synthesizing the control software for the industrial CPSs that are part of smart manufacturing systems, ultimately fostering their widespread adoption. The methodologies developed in DeFacto reason about systems using reliable compositional abstractions of system behaviors based on assume-guarantee (A/G) contracts. Intuitively, an A/G contract represents the interface of a component as a pair of assumptions and guarantees. Assumptions are the behaviors that a component expects from the environment; guarantees are the behaviors the component promises in the context of the assumptions. Contracts are then mathematical models that provide rigorous composition rules and mechanisms to analyze complex system behaviors, validate the design requirements, and develop system components in a modular and hierarchical way. The project has three main research objectives:
- Objective 1: Identify requirements and architectures for the design of industrial CPSs and define contract-based representations for the requirements and the system components at different abstraction levels.
- Objective 2: Develop a formal methodology, algorithms, and computational tools for contract-based requirement validation, design-space exploration, and model refinement. The methodology refines the system-level requirements and maps them to a system architecture and a set of control algorithms.
- Objective 3: Develop synthesis and mapping algorithms to generate software implementations from higher-level models of the system architecture and the control algorithms. The implementations include monitors detecting undesired behaviors due to uncertainty in sensors, communication networks, and machine learning components.
These opportunities will come, however, with a series of engineering challenges. A smart manufacturing system must integrate a diverse set of components while offering strong guarantees in terms of functionality, reliability, safety, and cost. This heterogeneity in components and system requirements inevitably calls for models, specification formalisms, and design constraints of different nature to represent a design space that is difficult to extensively explore in a reasonable time, an issue that is often recognized as the "explosion in complexity" of today's industrial CPS design.
Mirroring the success of electronic design automation (EDA) in taming the complexity of microchip design in the '80s, system design automation is expected to play a crucial role in reducing the complexity of CPS design. The scientific goal of DeFacto is to advance the state of the art in system design automation by developing novel modeling paradigms, scalable algorithms, and tools to aid the design of smart manufacturing systems. DeFacto aims at automatically defining the architecture (i.e. the system components and their interconnections) and synthesizing the control software for the industrial CPSs that are part of smart manufacturing systems, ultimately fostering their widespread adoption. The methodologies developed in DeFacto reason about systems using reliable compositional abstractions of system behaviors based on assume-guarantee (A/G) contracts. Intuitively, an A/G contract represents the interface of a component as a pair of assumptions and guarantees. Assumptions are the behaviors that a component expects from the environment; guarantees are the behaviors the component promises in the context of the assumptions. Contracts are then mathematical models that provide rigorous composition rules and mechanisms to analyze complex system behaviors, validate the design requirements, and develop system components in a modular and hierarchical way. The project has three main research objectives:
- Objective 1: Identify requirements and architectures for the design of industrial CPSs and define contract-based representations for the requirements and the system components at different abstraction levels.
- Objective 2: Develop a formal methodology, algorithms, and computational tools for contract-based requirement validation, design-space exploration, and model refinement. The methodology refines the system-level requirements and maps them to a system architecture and a set of control algorithms.
- Objective 3: Develop synthesis and mapping algorithms to generate software implementations from higher-level models of the system architecture and the control algorithms. The implementations include monitors detecting undesired behaviors due to uncertainty in sensors, communication networks, and machine learning components.
In the first phase, we analyzed the main requirements of manufacturing enterprises, focusing on small and medium enterprises. We categorized the requirements characterizing the Industry 4.0 paradigm, as well as the emerging Industry 5.0 trend. Then, we analyzed the status of manufacturing modeling tools and languages: we identified the modeling trends of traditional production systems. We investigated the scientific literature on the modeling of manufacturing systems. Our analysis highlighted that the modeling techniques currently used in the manufacturing field could not model many requirements of Industry 4.0 and all the requirements of Industry 5.0 systems. Thus, we searched for modeling languages able to capture the identified requirements.
SysML emerged from our analysis as the most promising modeling and design language for industrial CPSs. We identified the relations between the requirements for advanced manufacturing systems, the system formalisms involved in the requirements, and the SysML diagrams that can be used to model such formalisms. Then, we defined a methodology enabling the reuse of already existing system models and a SysML-based modeling strategy for system components and production recipes. The developed modeling strategy aims to create models supporting the most advanced functionalities required for manufacturing systems.
During the outgoing phase, at the Ming Hsieh Department of Electrical and Computer Engineering of the University of Southern California, we focused on the definition of methodologies and tools to provide formal support for the design of industrial CPSs. In particular, we implemented CHASE: a contract-based framework for the design space exploration, analysis, and optimization of CPSs. We developed methodologies for the validation and synthesis of discrete systems, the analysis of stochastic systems, and to perform architectural exploration. Then, we also tackled the problem of co-designing topology, scheduling, and planning of complex logistics systems, such as automated warehouses.
We developed a variation to the classic automation pyramid to collect the data necessary to support the design flow while enabling automation of existing systems. It automates the supervisory layer while improving data collection.
The results of the first two years were disseminated through articles at major international conferences and tutorials at major scientific venues in the field of cyber-physical systems, design automation, and manufacturing.
SysML emerged from our analysis as the most promising modeling and design language for industrial CPSs. We identified the relations between the requirements for advanced manufacturing systems, the system formalisms involved in the requirements, and the SysML diagrams that can be used to model such formalisms. Then, we defined a methodology enabling the reuse of already existing system models and a SysML-based modeling strategy for system components and production recipes. The developed modeling strategy aims to create models supporting the most advanced functionalities required for manufacturing systems.
During the outgoing phase, at the Ming Hsieh Department of Electrical and Computer Engineering of the University of Southern California, we focused on the definition of methodologies and tools to provide formal support for the design of industrial CPSs. In particular, we implemented CHASE: a contract-based framework for the design space exploration, analysis, and optimization of CPSs. We developed methodologies for the validation and synthesis of discrete systems, the analysis of stochastic systems, and to perform architectural exploration. Then, we also tackled the problem of co-designing topology, scheduling, and planning of complex logistics systems, such as automated warehouses.
We developed a variation to the classic automation pyramid to collect the data necessary to support the design flow while enabling automation of existing systems. It automates the supervisory layer while improving data collection.
The results of the first two years were disseminated through articles at major international conferences and tutorials at major scientific venues in the field of cyber-physical systems, design automation, and manufacturing.
This research improved the state of the art by identifying novel modeling and knowledge representation strategies to improve the design of industrial CPSs. It also contributed to improve the theory of assume-guarantee contracts: a very promising mathematical theory to reason about system design. Furthermore, the effort to develop CHASE is pushing toward the automation of the theory, thus making it more practical and usable for engineers and system designers. Finally, the automation software architecture developed in the project enables existing manufacturing systems to implement service-oriented manufacturing. Thus, making production systems more flexible and adaptable to market changes.