IoT Systems involve numerous interconnected things that sense or enact on the physical world to support customized software services for human beings. From a software and systems engineering point of view, such systems are essentially complex sociotechnical systems that lead to the development of dynamically adaptable, cyber-physical, systems. The adaptability with regards to the physical environment comes from a feedback (control) loop (e.g., MAPE-K loop) assimilating data from the sensors, building a model of the surrounding environment, planning or possibly predicting new scenarios, and soliciting the actuators accordingly, in the form of a sequence of actions. As with any sociotechnical systems, the planning process is usually semiautomatic, highly interacting with nal users to provide the best experience. Various software services have been developed in the past decade, leveraging important frameworks developed by the IoT community (e.g., protocols and gateways), and leading to a wide range of smart systems in energy, production systems, robotics, transportation, healthcare, agriculture among others. The smartness of the system comes from the ability to bring intelligence into the feedback loop. This intelligence primarily leverages the assimilation and curation of the acquired data. However, as a sociotechnical system, it is of outermost importance of also considering broader physical, economic, social and environmental concerns in which the systems and nal users involved. Since such information is di cult to get from sensors or to hard-code into the software itself, additional information must be combined with the available data to provide a holistic and systemic view of the system and its environment, support for making informed decisions. This need is currently supported by the concept of digital twins. When comes the time of designing such a feedback loop, modeling appears to be key. Modeling is key to capture any sort of knowledge in the form of descriptive models built from acquired observations or data, and modeling is also key to drive the development and evolution of complex ? The vision presented in this extended abstract is the result of the collaboration with many talented students and bright colleagues. I'm warmly thankful to all of them. Copyright c 2019 for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
systems in the form of prescriptive models reducing the accidental engineering complexity. The gap between the descriptive models and the prescriptive models can be made manually, or automatically through predictive models.
In this talk, I review the various types of models required for intelligently designing software services on top of IoT systems, and I discuss the different roles such models are playing in the overall lifecycle. I present the opportunities for the modeling community, as well as the open challenges to be tackled to achieve such a vision. In particular, I explore the required common modeling foundations for seamlessly combining the di erent types of models, and the development of complex digital twins to support informed decision making in the feedback loop of smart sociotechnical IoT systems. Benoit Combemale (http://combemale.fr) is Full Professor of Software Engineering at the University of Toulouse, and a Research Scientist at Inria. His research interests are in the eld of software engineering, including Model-Driven Engineering, Software Language Engineering and Validation & Veri cation; mostly in the context of (smart) CyberPhysical Systems and Internet of Things. He is also teaching worldwide in various engineering schools and universities. Prof. Combemale received his Habilitation in Computer Science from the University of Rennes 1 in 2015, and his Ph.D. in Computer Science from the University of Toulouse in 2008. Before joining the University of Toulouse, he was an Associate Professor at the University of Rennes 1, and has been visiting professor at McGill University and Colorado State University. Prof. Combemale coauthored 3 books, and 100+ journal and conference publications in the elds of software engineering. He also edited 2 books and various special issues in scienti c journals. He is chairing the Steering Committee of the conference series SLE, a deputy editor-in-chief of the journal JOT, and a member of the editorial boards of the journals SoSyM, COLA, and SCP. He has been the program chair of SLE 2014, ECMFA 2019 and ICT4S 2020, and general chair of MODELS 2016 and SLE 2017. He also serves as program committee member for various conferences and workshops in software engineering. Prof. Combemale coordinated and participated in many collaborative projects, and bilateral collaborations with industries. He is also a founding member of the GEMOC initiative, and currently lead the steering committee of the Eclipse Research Consortium GEMOC.