Modeling for Sustainability

Abstract

This report documents the program and the outcomes of Dagstuhl Seminar 18351 “Modeling for Sustainability”, August 26–31, 2018. Many different kinds of models, from engineering models to scientific models, have to be integrated and coordinated to support sustainability systems such as smart grid or cities, i.e., dynamically adaptable resource management systems that aim to improve the technoeconomic, social, and environmental dimensions of sustainability. Scientific models help understand sustainability concerns and evaluate alternatives, while engineering models support the development of sustainability systems. As the complexity of these systems increases, many challenges are posed to the computing disciplines to make data and modelbased analysis results more accessible as well as integrate scientific and engineering models while balancing trade-offs among varied stakeholders. This seminar explored the intrinsic nature of both scientific and engineering models, the underlying differences in their respective foundations, and the challenges related to their integration, evolution, analysis, and simulation including the exploration of what-if scenarios. Sustainability systems must provide facilities for the curation and monitoring of data sets and models and enable flexible (open) data and model integration, e.g., physical laws, scientific models, regulations and preferences, possibly coming from different technological foundations, abstractions, scale, technological spaces, and world views. This also includes the continuous, automated acquisition and analysis of new data sets, as well as automated export of data sets, scenarios, and decisions. The main function is to support the generation of what-if scenarios to project the effects on the different sustainability dimensions, and support the evaluation of externalities, especially for non rapidly renewable resources. Since the predictions are necessarily probabilistic, the system must be able to assess the uncertainty inherent in all its actions and provide suitable representations of uncertainty understandable by users. In addition to generating what-if scenarios to explore alternate model instantiations, the tool should be capable of generating suggestions for how to reach user-specified goals including quantifiable impacts and driving the dynamic adaptation of sustainability systems. These powerful services must be made accessible to the population at large, regardless of their individual situation, social status, and level of education. This seminar explored how Model-Driven Engineering (MDE) will help to develop such an approach, and in particular i) how modeling frameworks would support the integration of the various heterogeneous models, including both engineering and scientific models; ii) how domain specific languages (DSLs) would (a) support the required socio-technical coordination, i.e., engage engineers, scientists, decision makers, communities, and the general public; and (b) integrate analysis/probabilistic/ user models into the control loop of smart CPS (cyber physical system). DSLs are also supposed to provide the right interface (in terms of abstractions/ constructs) to be used as tools for discovering problems and evaluating ideas. The seminar served to identify critical disciplines and stakeholders to address MDE for sustainability and the research roadmap of the MDE community with regards to the development of sustainability systems. In particular, the seminar identified and explored four key areas: 1) research challenges relevant to modeling for sustainability (M4S); 2) a multidisciplinary collection of relevant literature to provide the foundation for exploring the research challenges; 3) three case studies from different application domains that provide a vehicle for illustrating the M4S challenges and for validating relevant research techniques; and 4) the human and social aspects of M4S. The cumulative results of the work performed at the seminar and subsequent collaborations will help to establish the required foundations for integrating engineering and scientific models, and to explore the required management facilities for evaluating what-if scenarios and driving adaptive systems. In addition, we envision to produce as an outcome of the seminar a representative case study that will be used by the community to assess and validate contributions in the field of modeling for sustainability.