Tasks, deliverables, and milestones
WP1 is organized around six tasks. All of these aim to explore and extend the state of the art in modeling and analysis of complex, large scale systems. Two of the tasks deal with the fundamentals of such methods. The remaining four deal with motivation and investigation of potential impact in the application domains and show cases considered in the project. The workflow is arranged so that the fundamental and applied tasks cross-fertilize one another. Fundamental tasks provide as input to applied tasks and overview of the types and capabilities of available methods, as well as novel developments during the life time of the project. The applied tasks on the other hand provide the motivation for the types of methods that need to be investigated further and study the implications of the methods on the specific application domains.
The six tasks are as follows:
In traditional system theory, system identification is associated with methods for using observations of the system evolution to estimate the values of parameters that cannot be measured directly. Typically such parameter identification methods rely on the hypothesis that the structure of the system dynamics is known. In complex, networked control systems, however, this hypothesis is usually not realistic. For example, in the study of biological networks it is very often the case that not only the parameters of the different chemical reactions (reaction rates, binding propensities of proteins to the DNA, production rates of proteins, etc.) are unknown, but also the structure of the network itself (how the different chemical species involved in the reactions affect each other) is uncertain; in fact it is very often the case that one does not even know which, or how many species are involved in the process of interest. As another example, the identification of models at a discrete event level of abstraction it is often based on samples of the system behavior in terms of sequences of events, without any knowledge of its internal structure. To be able to deal with problems like these, novel methods are needed that can cope with the joint parameter and structure identification of systems, based on data on the system behaviors. Similarly, filtering methods are traditionally used for on-line estimation of the state of the system from partial measurements that are corrupted by noise. Over the years powerful filtering methods have been developed. Some exploit system structure to obtain tractable state estimates, while others aim for generality and rely on computationally intensive Monte-Carlo estimation. None of these methods, however, is capable of dealing with the complexity of networked systems. For example, in the area of power transmission and distribution networks, state estimation based on optimization is routinely used to provide information on unmeasured quantities such as power flows, voltages, phase angles, etc. The performance of such methods heavily depends on knowledge of the topology and parameters of the underlying power transmission or distribution network: Connectivity of the transmission lines, line impedance values, etc. These quantities, however, are rarely known precisely and often change on-line without being directly detected. This is especially true in the European energy market, which is both fragmented across several countries and mostly liberalized, with different operators dealing with different parts of the network; in this situation each country and operator naturally gets only a partial view of the overall picture. Novel state estimation methods are therefore needed to detect and keep track of such structural changes. Related to this is the problem of fault detection and isolation. Also in this case, faults are not immediately detected, it is crucial, however, to determine their presence and estimate their effects from indirect measurements if the safe operation of the system is to be assured.
Complex systems have a tendency to surprise engineers with unexpected behavior. This is true not only for biological systems (where understanding the sources of emergent behavior is essentially the goal) but often also for engineering systems. Because of their hierarchical structure, complex engineering systems involve interactions at many levels. In road transportation systems, for example, traffic patterns at a regional and national network level “arise” out of distributed local interactions between vehicles and drivers. Understanding how the overall behavior emerges out of the local interactions is a challenging task that cannot be addressed by state of the art system theory methods. The problem is complicated further when one intends to intervene on the process. Control decisions often intend to address requirements at the higher levels but can only be implemented at the lower levels. Understanding the impact that the control commands will have on the emergent behavior of the system as they filter down the hierarchy is essential for the successful implementation of control policies. Successful examples of this principle can be found in the area of communication networks and manufacturing. Here fluid limit approximations are used to design stable scheduling policies at the network level. These policies are then implemented at the “packet” or “job” level in a way that ensures that the stability properties of the fluid limit are preserved. There is a growing need for the development of methods that can deal with such situations in many different settings; after all, understanding and analyzing how emergent behavior arises out of distributed interactions is a common theme of many complex systems, including transportation, biological systems, financial systems, etc.
This task will investigate the needs of problems in transportation in terms of modeling and analysis methods, as well as the potential of the applications of such methods in transportation problems. The task will start with a wide-ranging study of the types of models, levels of abstraction and analysis methods that arise in transportation. Special emphasis will be placed on issues such as abstraction and hierarchical requirements (moving, for example, from the level of individual vehicles or aircraft to flows in a highway network or the air traffic management system). State estimation and parameter identification tasks will also be investigated; for example estimating the state of a model of traffic flow from sparse, distributed camera, and/or cell-phone, and/or magnetic sensor data. The results of this study will be paired with the outcome of the state of the art studies of Tasks 1.1 and 1.2 to identify synergies and limitations of the existing methods. Novel methods developed under tasks 1.1 and 1.2 to circumvent these limitations will be assessed as part of Task 1.3.
This task will investigate the modeling and analysis needs that arise in energy problems. The task will again start with a wide-ranging study of the types of models, levels of abstraction and analysis methods that arise in this domain. Issues that will be addressed include, among others, the use of hybrid models in power networks to capture the coupling between continuous quantities (voltages, phase angles, frequencies) with the changes in the discrete network topology, fault detection and isolation methods for power systems, stochasticity in the energy supply arising from the introduction of renewables, dynamic state estimation and topology identification in power networks, complexity in nonhomogeneous transport phenomena captured by partial differential equations, etc.. The results of this study will again be paired with the outcome of the state of the art studies of Tasks 1.1 and 1.2 to identify synergies and limitations of the existing methods. Novel methods developed under the fundamental tasks to circumvent these limitations will be assessed as part of Task 1.4.
This task will investigate the modeling and analysis needs that arise in biomedical applications. Issues that will be addressed include, among others, structured modeling for biological systems, hierarchies and emergent behaviors, the interplay of robustness and network topology in biochemical networks, distributed control of cell populations, and complexity-aware data-based modeling. The results of this study will again be paired with the outcome of the state of the art studies of Tasks 1.1 and 1.2 to identify synergies and limitations of the existing methods and assess the potential impact of novel methods.
During the last year of the project a small task will be devoted to exploring the modeling needs of emerging application domains not considered during the main part of the project. Such application domains include (but will not be limited to) finance and insurance, macro-economic models and modeling of human-in-the-loop systems (where the human is e.g. an operator in a power plant or an air traffic controller). As part of this task, we will take a holistic view and attempt to identify future challenges in the modeling an analysis of complex large scale systems, including but also going beyond the ones identified in Task 1.3-1.5.
To read the available deliverables you must be a Level 1 Partner.
To upload a deliverable you must be a WP Leader.
Please log in to have access to deliverables and to download the template.