DDDAS: Dynamic Data Driven Applications Systems

APPLICATION EXAMPLES

List is meant to be indicative of examples and areas of interest, and not limiting. Other, novel applications are highly encouraged

• Generalized methodology for state estimation and prediction. State estimation and prediction for dynamical systems is accomplished by blending a simulation model, executing in real time, with measured data - predictions are based on the simulation and these predictions are corrected as data become available. The most notable example of this predictor-corrector approach is the celebrated Kalman Filter. DDDAS has the potential to enable and improve state estimation and prediction for complex dynamical systems operating in uncertain environments, where all the data required for the simulation are not known apriori. DDDAS may be utilized to determine when and where in the simulation model, data should be injected to correct predictions with bounded computational resources. DDDAS may also be used to take advantage of information in the predictions to suggest in real-time how to adapt the sensing strategy to trade accuracy and prediction horizon against resources in the face of uncertainty. The utilization of the DDDAS paradigm in estimation and prediction can have a big impact in many areas, from improving engineering systems such as Advanced Driving Assistance Systems for automobiles and tracking algorithms for Air Traffic Control, to neurobiological and bio-molecular systems, to enhancing oil exploration methods and capabilities, to nanometrology.
  
  
• Advances in sensor technology are making it possible to have manufacturing supply chains improve visibility of products and processes between different parts of the chain. For instance, the inclusion of RFID tags on parts for work-in-process inventory allows faster updates of system status, resulting in the possibility of real-time re-assignments that will improve the performance of a manufacturing system.
  
  
• Virtual operations re-planning and control situations occur regularly in manufacturing and service systems subject to unpredictable disruptions such as equipment breakdowns. Event-driven simulations running in parallel to current operations offer the opportunity to rapidly adjust to such disruptions and select appropriate recourse. Virtual operations simulations that constantly update the system state have been in use for some time, but DDDAS would significantly enhance their capabilities by incorporating diagnosis elements allowing the simulation to determine autonomously what new data is required to define the nature of a disruption and how to provide effective simulated analysis of alternative courses of action.
  
  
• Earthquake tolerant buildings and bridges. DDDAS capabilities for controlling dampers in the support columns and joints.
  
  
• Fire propagation prediction and management. In open spaces (forests, for evacuation, containment, distribution of fire personnel) and in closed spaces (large buildings for evacuation planning and action, containment to mitigate damage; in submarines for containment to mitigate damage to people and structure), and for decision support systems to minimize damage casualties.
  
  
• Advanced Driving Assistance system: The driving assistance system continuously predicts the vehicle motion by model-based real time simulation which incorporates measurement data such as the steering angle, vehicle accelerations and yaw rate. Vision may be another useful input. Accurate prediction of vehicle motion by the system is critical to assist the driver during normal driving and emergency situations. Handling emergency is particularly challenging. Sudden changes of the vehicle states such as a bursting tire drastically change the vehicle dynamics. Tire bursts may be detected by measurements from sensors such as tire pressure sensors and a roll sensor, and these measurements must be incorporated in the selection of a right model and model parameters for accurate prediction of vehicle motion to maintain vehicle stability during emergency. This is an example of fault detection and fault management, and similar situations take place in many engineering problems.
  
  
• DDDAS capabilities to enhance chemical process design and control. Specifically, design methodology that moves towards more and more all-encompassing computer aided designs of whole enterprises incorporating availability of a variety of raw materials, effect of ambient conditions on plant operations, moment-by-moment variations in the markets for manufactured products, as well flexibility in equipment usage for varying product lines, continuous and batch processes, etc. The incorporation of control considerations during the original design stage could is also relevant.
  
  
• Real time control of complex thermal-fluid processes. This is difficult process to manage, because the simulations needed cannot be run fast enough to interface with a control algorithm that can effect changes in operating conditions. An example is machine tool compensation for thermal expansion errors in precision machining. DDDAS will permit real-time control of complex thermal-fluid processes, improving the coupling of continuous monitoring of data with robust simulations dynamically incorporating changing geometrical, environmental, or operating conditions. Successful implementation of DDDAS could allow measurements data to be supplied dynamically into a more robust modeling and control algorithm, allowing it to run in real time.
  
  
• Biological 'real time' experiments can involve time scales ranging from nanoseconds for molecular and sub-molecular dynamic processes to years and decades for ecological changes. For example, in experiments recording neural activity, where the dynamics occur at different speeds, it could be very useful in general to have 'smart devices' that would shape an experiment in real time. If a dynamic, data-driven system were available, an 'on-line comparator' device could track the data stream as it comes in, put it in the context of related data or experiments, and help to make a decision about what measurements to perform next. In addition recent demonstrations of grid computational approaches to bio-structural molecular dynamics opens long-scale temporal domains to the theoretical analysis of complex biological systems.
  
  
• Next Generation Micro-array Management System provides methodology for efficient measurements on plant genetics.
  
  
• Biological systems, which are being defined at increasingly comprehensive levels of detail, provide another context within which to consider the impact of the DDDAS paradigm. Models necessary for understanding such large-scale and complex systems, involving as they do multiple interacting subsystems, clearly require DDDAS capabilities, particularly since the hierarchical complexity of these models must adapt to the data stream, since the system's internal and external interactions will change as the state of the system changes. As the complexity of biological systems, from cellular to population levels becomes the focus of inquiry, DDDAS capabilities will become integral to bioscience investigations.
  
  
• Integrated Image-Guided Interventions - Currently, the technology employed in today’s operating rooms does not accommodate the real-time, three-dimensional (3D) imaging needs of surgeons. Most imaging devices display anatomic information in only two dimensions (2D), making navigation to the desired site through small incisions extremely difficult. This is because the surgeon's direct line of sight is disrupted, resulting in a "visual disconnect" from the operative target of interest. For 3D views there are a variety of imaging devices which can guide minimally invasive therapies by fusing images from different imaging devices, such as MRI and CT. These fused images allow the surgeon to have "x-ray vision," or to see through solid organs in order to reach the target area with minimal damage to the surrounding organs. Furthermore, integrating 3D images with semi-autonomous devices, such as surgical robots, will enable surgeons to perform minimally invasive procedures in a manner never before experienced. Also critical to the success of image-guided interventions is the real-time integration of patient information with the images being generated during the surgical procedure or treatment. Images, as well as critical information on the patient’s current status and history, could be provided in an easy-to-understand display. Thus, all relevant information could be provided at the point of patient care.
  
  
• The biodiversity and bio-complexity of the world's terrestrial and aquatic communities and ecosystems are subject to dramatic changes due to habitat transformation, invasions of exotic species, chemical contamination, diseases and epidemics, climate change, and floods and drought. Studies into the change must look beyond isolated events, and analyze the complex interdependency and interaction in the way multiple factors impact the ecosystem and cause these changes. Static models are inadequate in capturing the contributing ensemble of factors. To analyze the complex causes and to gain insights into such processes, models are needed that can integrate data from several scales of observation, such as molecular, individual, population, and regional models of analysis. The enhanced integrated modeling would need to take advantage of real time measurement methods, including remote sensing and geographic information systems.
  
  
• Hydro-complexity – Weather, Water and Pollution: Weather processes are modeled at large spatial scales; hydrologic processes are modeled at smaller spatial scales, and groundwater pollution is modeled at even smaller spatial scales and much longer time scales. One of the greatest challenges in "water" process simulation is to find a practical way to connect models that operate and require inputs of dynamic data at vastly different scales. By rising to the challenge of interconnecting models with large scale diversity, the simulations would be better able to forecast movements of extreme storms such as hurricanes or tornados (as happened so recently in the Midwest), cloud-burst flooding moving downstream such as that which disrupted Colorado State Univ. a few years back, spills of toxic materials into rivers or other natural water bodies, and the generation of landslides and debris flows from wetted hill slopes, a widespread problem in the western states. DDDAS offers vast opportunities for addressing non-linearities in this highly interactive system. One specific need here is to use the dynamic data to adjust the system of equations when processes or interactions among processes change over time or vary over space.
  
  
• Design and configuration methodologies for sensor networks and traditional networking systems. As the size, capability and complexity of the networking systems grow, achieving a global, multi-scale, and multi-layer understanding of complex large-scale networks is imperative for the successful design and development of robust, secure and scalable next-generation Internet protocols. Ubiquitous sensor networks pose marked challenges as to what the configuration and architecture of the system of sensors will be, and in deciding what kinds of data to collect and to deliver to the application. The DDDAS paradigm offers a unique opportunity to study the dynamics of the global-scale heterogeneous networks and understand the behavior of the associated infrastructure.
  
  
• In Geosciences, simulations of processes operating in natural (as opposed to engineered) systems can be made dynamic by introducing observations in real time. In a very quick time frame, simulation models would be better able to forecast movements of extreme storms such as hurricanes or tornados (as happened so recently in the Midwest), cloud-burst flooding moving downstream such as that which disrupted Colorado State Univ. a few years back, spills of toxic materials into rivers or other natural water bodies, and the generation of landslides and debris flows as hill-slopes become wetter, a widespread problem in the western states. In a time frame that is a bit slower, the improved simulation models would be made better able to track the spread of pollutants in groundwater, to forecast drought patterns in time and space (for example during El Nino periods), snow-melting to generate floods (more quickly) or provide water to downstream users (less quickly), or protect aquatic environments from climate variations or anthropogenic activity. In a much slower time frame, hydrologists are beginning to work with scientists in geomorphology on a "Community Sediment Model" that simulates how rivers erode and leave sedimentation. Over time, the water deposits leave an alluvium whose properties are needed for groundwater management and for the design of foundations for engineering structures. If we can model, the processes creating alluvium over geologic time, we will have a much stronger quantitative basis for defining soils and geology beneath us. Each new well or other hole would provide additional data for model refinement. Other applications are found in responses of the Earth surface to climate change or to long-term alterations of anthropogenic activity.
  
  
• Even greater opportunities for DDDAS in addressing non-linearities in systems. The next step past adjusting the simulations when getting new data is to adjust the system of equations when processes or interactions among processes change. For example, if the simulation is for river flooding. One has a system when the floods are contained in the rivers, another set of circumstances when levees are overtopping, and still another set when the water is spreading and then draining over a large floodplain. The modeling system must be shifted and the shifts (nature of the revised equations) are location dependent. The same thing happens when hill-slope soils change from wetting to moving (and then stopping).
  
  
• The oceanographic community at large has interests in DDDAS in order to help optimize observing systems for important scientific studies. For example to support environments like the ones envision in the pilot project described below and reduce the oversampling resulting from the current approaches. The pilot project will use the approach of integrating coupled observational / modeling systems. This experiment will use AOSN observational assets developed under ONR funding (gliders, profiling floats, AUV’s, remote sensing, ship-board observations, coastal radars and buoys), it will employ multiple oceanographic modeling/assimilation schemes, and it will use control theory and dynamical systems theory to optimize sampling strategy in order to optimally observe and predict the features of an upwelling event. The pilot’s purpose is to work on the underlying theories and structure to bring these elements together into a functional system. The project will over sample the oceanographic field in order to support verification of the skill of the system. This is an actual field exercise that will occur late this summer.
  
  
• Within the geosciences other examples of benefit from the DDDAS framework include the following: the topic of fusing data with models has a long history and the output from model-data fusion have been used as to analyze systems, to forecast system behavior, to design observing systems, to modify the behavior of sensors in sensor networks in order to respond to transient events, and to drive adaptive sampling protocols for real-time assimilative prediction systems. Examples of past and current work in this area include: the development of many schemes for numerical weather prediction in research and forecast modes; experiments such as THORPEX, FASTEX, NORPEX and WSR that look at the effects of using adaptive meteorological sampling to improve weather forecasting; ocean state estimation; autonomous control systems for ocean gliders and AUVs; adaptive ocean sampling and observing system simulation experiments (OSSE’s) such as those carried out at LEO-15, in Cape Cod Bay, within the Poseidon project, and as part of a number of projects in this area funded or planned by the Office of Naval Research; and data assimilation in atmospheric chemical transport modeling. Related work includes the use of inverse methods to determine: atmospheric and oceanic tracer distributions, including sources and sinks of tracers; the structural robustness of marine ecosystem models; the structure of the Earth’s interior; three-dimensional information about the Earth’s atmosphere and ocean from radio occultation data and acoustic data, respectively; analysis of the Earth’s magnetic field. This type of work with inverse models can yield useful information for the designers of observing systems, especially observing systems intended for use with data-assimilative models and objective analysis systems.
  
  
• Further research activities that also can benefit from the DDDAS concept are those anticipated in areas such as: practical data assimilation methods for physical, chemical and biogeochemical models for the atmosphere, ocean, hydrologic, geophysical and other environmental systems, with and without adaptive sampling; incorporating data assimilation into magnetospheric and ionospheric models, both for state estimation and for research into forecasting “space weather”; adaptive and event-driven sampling schemes and/or control schemes for resource-limited environmental sensing networks, where the resource limitations may be associated with factors such as power, communications, data storage capacity, or real-time analysis capabilities; data-driven control schemes for autonomous sensor platforms with a range of different power budgets (for example, AUVs and UAVs); incorporating data assimilation into climate and forest fire models; and OSSE research for oceanic, atmospheric, hydrological, geophysical and environmental systems, including observation and modeling systems design for natural hazard assessment.
  
  
• Design and configuration methodologies for sensor networks and traditional networking systems. As the size, capability and complexity of the networking systems grow, achieving a global, multi-scale, and multi-layer understanding of complex large-scale networks is imperative for the successful design and development of robust, secure and scalable next-generation Internet protocols. Ubiquitous sensor networks pose marked challenges as to what the configuration and architecture of the system of sensors will be, and in deciding what kinds of data to collect and to deliver to the application. The DDDAS paradigm offers a unique opportunity to study the dynamics of the global-scale heterogeneous networks and understand the behavior of the associated infrastructure.
  
  
• Modeling of Networks: DDDAS goes beyond the traditional event-driven network simulation paradigm to one that can dynamically incorporate on-line and archived measurements from real networks to allow a better understanding and a more accurate prediction of the network behavior for a wide range of time-scales, a broad spectrum of spatial network topology structures, and multiple protocols interacting with one another and across the different networking layers, especially as one goes from wired to wireless and streaming data from networked sensor systems. Research issues which need to be addressed in order to take full advantage of DDDAS and achieve a deep and solid understanding of the behavior of large-scale networks: Development of new simulation models which have the capability of handling dynamically injected data; Development of new algorithms and schemes that can identify and extract, within the context of the simulated network, meaningful patterns from the on-line data and use these patterns to guide the network simulation toward accurate understanding of the effective network behavior; Development of new algorithms to dynamically characterize different states or behaviors of the network, and use this characterization to achieve different levels of simulation granularities commensurate with the objectives of the simulation; Development of novel, data-driven network monitoring, measurement and analysis techniques in support of a new generation of network simulation tools for future large-scale network architectures.
  
  
• Large research facilities such as the Spallation Neutron Sources (SNS), and current and future synchrotron light sources will benefit from the DDDAS. The SNS for example will produce data on the positions and motions of atoms and spins in materials with unprecedented detail. DDDAS could help produce new software to analyze the data from these and other large facilities and enable the new types of science that are possible with modern these modern instruments. NSF currently supports a Conceptual Design Engineering project of software for a distributed data analysis for neutron scattering experiments. Construction proposal could be supported through DDDAS. Real time control of charged particle accelerator systems: Accelerators now in design for many uses from bioscience to nanotechnology have very demanding requirements such as holding beams on target with accuracies of microns or even nanometers. Achieving this requires extremely precise control of hundreds of parameters. Implementation of DDDAS in the control systems of such apparatus will allow use of the beam parameters themselves for design of any needed corrective actions (parameter adjustments) in response to earth motions, temperature changes, power system drifts and the like.
  
  
• The National Science Foundation partially supports two particle physics experiments that will collect data at the Large Hadron Collider (LHC) that is being constructed at CERN, located near Geneva, Switzerland. This facility will consist of a superconducting particle accelerator providing two, counter-rotating beams of protons, each beam having an energy up to 7 TeV (1 TeV = 1012 electron volts). The two experiments, A Toroidal LHC Apparatus (ATLAS) and the Compact Muon Solenoid (CMS), are being constructed to characterize the different reaction products produced in the very high-energy proton-proton collisions that will occur in intersection regions where the two beams are brought together. The principal scientific goal for the LHC is to enable searches for the Higgs particle, the existence and properties of which will provide a deeper understanding of the origin of mass of the known elementary particles, and for particles predicted by a powerful theoretical framework known as supersymmetry. It will also enable the investigation of the possibility that there are extra-dimensions in the structure of the universe. A total of 34 international funding agencies participate in the ATLAS detector project and 36 in the CMS detector project. Both the NSF (for $81M) and DOE ($250M) are providing U.S. support to the detector construction for these experiments, while DOE is also providing some support ($200M) for the accelerator itself. The LHC is expected to produce initial collisions in 2007. The U.S. LHC Collaboration has been a leader in the development of Grid-based computing. The Grid will enable the enhanced participation of universities and thus the training of students, in both state of the art science and computational techniques, in a project that is centered overseas. The Grid is expected to have a broad application throughout the scientific and engineering communities.
  
  
• In the social and behavioral sciences DDDAS can enable real time adaptive and dynamic approaches to interviewing, cognitive measurement and experimentation, public and private decision-making, etc. Applications abound. One illustrative example is the dynamic adaptation of data collection systems and programmatic response (e.g. for police patrols, health services delivery, emergency evacuation and news reporting for natural hazards, vote counting and election oversight, and traffic management) that relies on fundamental understanding of cognition, learning processes, small and large-group behavior, demographics, and crisis decision-making.