Interactive and Optimal Configuration of Cyber Physical System Product Lines

Nowadays, Cyber Physical Systems (CPSs) are everywhere in our daily life from video conferencing systems, subsea oil and gas production systems, smart buildings, smart healthcare products. These systems are mostly developed by reusing existing system designs, instead of from scratch, for the purpose of reducing development cost and improving product quality. Systematically doing so follows into the research stream of system product line engineering. However, configuring large-scale CPS product lines requires a systematic, interactive and maximally automated methodology (with tool support). The goal of the Zen-Configurator project is to increase the cost-effectiveness of configuring large-scale CPS product lines at different phases of the development lifecycle of such systems (e.g., pre-deployment, post-deployment, and runtime operation phases). To achieve this goal, we maximally automate error-prone and costly manual configuration activities and optimally assist the interactive configuration process. On one hand, the project relies on advanced technologies of constraint solving/evaluation, optimization using search algorithms, machine learning techniques and propose state-of-art algorithms to enable automated configuration activities. On the other hand, the project grounds itself to address real challenges faced by industry and propose a practical and applicable solution and apply it to solve real-world problems. In the context of the project, we have developed a methodology with tool support to address pre-deployment and post-deployment phase configurations of CPSs.

In this project, we proposed an interactive and optimal configuration solution (with tool support named as Zen-Configurator) with three key functionalities implemented: Decision Inference, Decision Ordering and Consistency Checking. Zen-Configurator is based on the theoretical foundation including: 1) cost-effective optimisation for supporting decision ordering, 2) optimised constraint solving to enable decision inference, 3) effective formalisation of consistency checking, and 4) novel algorithms to enable the three functionalities with user acceptable performance.

The current practice of manually configuring any non-trivial Cyber Physical Systems (CPSs) product line is often error-prone and labor-intensive. In other words, the quality and productivity of the product configuration process cannot be ensured. In critical domains (e.g., oil and gas), poor quality and low efficient services provided by CPSs (e.g., subsea production systems) will cause severe issues (e.g., oil or gas leaks to the sea). In this project, we aim to improve the quality and productivity of configuring large-scale CPSs. Correctly configured products in CPSs are often directly deployable and operational systems. The method we take is to apply Model Based Engineering (MBE) technologies to automate the configuration process of Product Line Engineering (PLE) to the maximum extent. Effectiveness of a PLE approach for CPSs is characterised by its support for abstraction and automation. Abstraction plays a central role in software reuse, which is required to capture all relevant information in a concise and expressive manner to support automated configuration of products. Automation is required for effective selection and customisation of reusable components. We will propose in this project an interactive and optimal configuration solution (with tool support named as Zen-Configurator) with three key functionalities implemented: Decision Inference, Decision Ordering and Consistency Checking. In literature, it does not exist a configuration solution that supports all these functionalities in an integrated manner with a user acceptable performance. To address this R&D challenge, we will base our solution on the theoretical foundation including: 1) cost-effective optimisation for supporting decision ordering, 2) optimised constraint solving to enable decision inference, 3) effective formalisation of consistency checking, and 4) novel algorithms to enable the three functionalities with user acceptable performance.