This high-risk, high-reward project is concerned with environmental-impact modeling for infrastructure-network decision-support. The intellectual merit of the research lies in the introduction of an entirely new modeling framework which broadly captures environmental impact in infrastructures by meshing computing and controls concepts, and in the development of a suite of analysis and design tools for this framework. The proposed modeling framework draws on the influence model, a structured representation of interacting Markov chains that can capture the intricate spatio-temporal patterns that are observed in environmental impact, and yet admits rapid computation of local statistics due to its moment-linearity characteristic. By building on the influence-modeling construct, the PIs expect to develop a hybrid model for environmental-impact evolution. A second key task is to develop a suite of tools that allow the use of the environmental-impact model in decision support, with a particular focus on inference and analysis tools.
The broader impact of the proposed research is twofold. First, the research has the potential to impact computation and communication technologies in numerous widely-used infrastructures. Second, we the research is envisioned to enact a paradigm shift in modeling for large-scale computation, toward models that incorporate control-theoretic constructs to permit analysis and design of network (and multi-network) dynamics. Education and research-dissemination activities are planned to foster both the societal and computer-science-related impacts of the work. Highlights include multi-disciplinary course-development efforts, and industry-interface and research-experience provisions for undergraduates.
This high-risk, high-reward project is concerned with environmental-impact modeling for infrastructure-network decision-support. The intellectual merit of the research lies in the introduction of new modeling frameworks which broadly capture environmental impact in infrastructures by meshing computing and controls concepts, and in the development of a suite of analysis and design tools for this framework. The results from this project have been documented in one book chapter, nine journal publications, five student theses, and over twenty conference publications. Specific outcomes include a modeling framework that draws on the influence model, a structured representation of interacting Markov chains that can capture the intricate spatiotemporal patterns that are observed in environmental impact, and yet admits rapid computation of local statistics due to its moment-linearity characteristic. The PIs have used the modeling framework to i) capture uncertain weather propagation, a key factor that complicates strategic air traffic management, and ii) partition autonomous robot teams in a distributed fashion. Additional modeling outcomes include the first airborne random mobility models in the literature that are realistic to capture stochastic smooth aerial mobility and are simple enough for tractability. The stochastic mobility models build the theoretical foundation for the evaluation and design of networking strategies. Moreover, estimation of network models from noisy data has also been pursued with major results on linking network topology with estimation performance. By building on the modeling construct, the PIs have developed an integrated framework for environmental-impact evaluation, using the analytical jump-linear approach and the effective probabilistic collocation effective simulation approach. The jump-linear approach uses stochastic approximation to achieve a tractable analysis. The probabilistic collocation method (PCM) smartly chooses a few simulation points to produce a low-order mapping between uncertain parameters and system output, which predicts the correct output statistics. Multivariate PCM has been developed, together with the use of it for uncertain evaluation and optimal decision-making. Educational and outreach outcomes include the i) training of eight graduate students and multiple undergraduate students in stochastic system modeling and decision-making research, ii) development of a new graduate-level course Systems Modeling and Simulation, which advocates interdisciplinary research, and has attracted students from all engineering departments, iii) outreach to K-12 teachers in a summer RET program, and iv) outreach to the summer undergraduate program in engineering research (SUPER) at UNT. The broader impact of the proposed research is twofold. First, the research has the potential to impact computation and communication technologies in numerous widely-used infrastructures. Second, the research is envisioned to enact a paradigm shift in modeling for large-scale computation, toward models that incorporate control-theoretic constructs to permit analysis and design of network (and multi-network) dynamics. Research dissemination activities include invited talks, conference presentations, outreach activities, website construction, which foster both the societal and computer-science-related impacts of the work. Highlights include multi-disciplinary course-development efforts, and industry-interface and research-experience provisions for undergraduates.
