Designing an active recommender framework to support the development of reasoning mechanisms for smart cyber-physical systems

Our research concentrated on a newly emerged problem related to smart cyber-physical systems (S-CPSs). The essence of the problem is that S-CPSs are based on reasoning mechanisms that enable them to generate context-dependent solutions for various application problems. Typically, development of application-specific reasoning mechanisms (ASRMs) is a challenge for all stakeholders including software developers, knowledge engineers, and application developers. The research was conducted based on the global research hypothesis stated that an active recommendation framework (ARF) is a computer-aided design support tool with new features, including the coupling of process monitoring and decision support functionalities for compositional design of ASRMs of S-CPSs is not yet proposed. The whole research project was methodologically framed by a logical flow of four research cycles. Each research cycle addressed different aspects of the ARF development. The selected application context was a specific part of the design process of an automated parking assist system (APAS), as the target ASRMs. The research activities included: (i) knowledge aggregation, demarcation of the domain of interest, and specification of requirements; (ii) functional and architectural conceptualization of the ARF; (iii) computational implementation and operationalization of the demonstrative modules; and (iv) validation of the usefulness of the recommendations generated by the implemented demonstrative modules of the ARF. Initially proposed by researchers of the hosting Section of Cyber-Physical Systems Design, the concept of an ARF as significant novelty and supposed to play an influential role in the future. The term “framework” was used to refer to a purposeful enabler that arranges and rationalizes design activities, information processing, and designer-system interaction. The term “recommender” expresses that, as a complex system, the ARF derives context-dependent advice for the designer based on a comprehensive system model of the concerned (specific) design process. The term “active” refers to the fact that the ARF continuously monitors the design process and spontaneously interacts with the designer wherever it is needed in the design process. Our conclusions have been that the ARF goes well beyond the concepts of traditional SEFs and static recommender systems.