ARC Complex Open Systems Research Network

COSNet: ARC Complex Open Systems Research Network

characterising and analysing complex systems for explanation, prediction and control

excerpt from original ARC funding proposal

1.1 Summary and objectives
1.2 Networking strategies and programs

1.2.1 COSNet Nodes
1.2.2 CSIRO Centre for Complex Systems Science (CCCSS)
1.2.3 Technology and Tool Transfer
1.2.4 Summer Schools
1.2.5 Joint Activities and Sponsorship
1.2.6 Experiment and Observation
1.2.7 High-Performance Computation and Visualisation

The objectives of the Complex Open Systems Research Network (COSNet) are to

Over the past decade complexity has been recognised as the common frontier in the physical, biological and social sciences. This Network will be open and inclusive and will link specialists in all of these sciences through five generic conceptual and mathematical themes

The Network will assemble the capability to address some hard and up-to-now intractable questions, such as

These, and many others, are the kinds of big questions that complexity theory seeks to answer by providing the conceptual and mathematical tools to describe and model open, non-equilibrium systems with many interacting components. By preparing Australian researchers and graduate students to move beyond the comfort zone of their own specialisations, to learn a language in which they can communicate, the Network will provide an integrative force that will foster new industries and lead to benefits in human health, national security, economic wellbeing and sustainability.

Some background to this proposal is given at 3.7 – it is built on over a decade of preparatory work.

Back to Top

To catalyse the interaction between researchers in geographical locations and disciplinary areas, COSNet will be built around five General Themes that link many Application Areas (see 1.3).

We plan a variety of interaction tasks—technology and software/analysis tool transfer, e-collaboration, and workshops—with the aim of stimulating dialogue and focussing on the deeper conceptual problems of complexity.

We propose to manage our interaction tasks collaboratively with the CSIRO Centre for Complex Systems Science (CCCSS), who have a similar strategy, allowing a deeper interaction between pure and applied research. Where appropriate, COSNet will integrate with existing Working Groups that are already active within CCCSS, expanding their academic component.

There are currently three such groups active (see http://www.dar.csiro.au/css/wgroup.shtm):

It will also energise other planned, but not yet active, CCCSS working groups, in particular the Dynamical Systems Working Group.

Research Workshops

A key interaction mechanism for COSNet will be short research workshops held throughout each year at ANU or at the nodes. Most workshops will be focussed on the Application Areas described in 1.3 and organised by coordinators expert in these areas, convening an organisational committee with a range of cross-disciplinary expertise. In the Call for Proposals, issued each year, organisers will be encouraged to involve participants with a spectrum of interests and backgrounds. Early career and female organisers and participants will be particularly encouraged. Integration with the programs of the CCCSS, ACCS and MASCOS will be desirable, so as to leverage the considerable investment already made in setting up these centres.

Extended Research Workshops

To foster the development of general complex systems theory and methods, there will also be an extended research workshop each year devoted to one of the General Themes described in 1.3, running for two or more weeks so as to allow time for extended discussions and actual research during the workshop.

Remote Conferencing

The Access Grid facilities available already at many sites will be utilised by the general Working Groups and others. COSNet will also assist participants with low-cost facilities for their personal computers, such as Web cameras, so teleconferencing can become a convenient and regularly used way for the Working Groups to function.

Back to Top

The nodes listed in D3 have agreed to host Network activities such as workshops, host Network visitors, provide teleconferencing facilities, run an institutional complex systems seminar series and maintain a web page linked into the COSNet web site.

Back to Top

The CSIRO Centre for Complex Systems Science is a key partner in this application, and the eight CSIRO Participants named at A3, including its Director, Dr John Finnigan, will participate in the Network through this Centre. CCCSS is a distributed Centre comprising a small core group in Canberra that manages the Centre's activities and performs basic research in complexity, and a set of projects in CSIRO Divisions. These projects span areas from industrial physics to economics and social science, and many of them have university partners. The philosophy behind this arrangement grew from the observation that many of the advances in complex system science over the last decade have occurred at the interfaces between traditional disciplines or areas of application. By developing CSS projects in a wide variety of CSIRO Divisional contexts the CCCSS exploits a rich variety of problems and ensures a ready application for theoretical advances. At the same time it devotes considerable effort to linkage and transfer of ideas between the various projects by managed network activities. One of 5 areas of 'Emerging Science' identified in CSIRO over two years ago, the CCCSS has a current budget of around $5.5M pa and devotes over $0.5m pa to linkage and networking activities. Dr Finnigan proposes to manage CCCSS linkage processes cooperatively with COSNet, so that the networking activities of the ARC Network and that of the CCCSS can be seamlessly integrated and jointly funded (on a "no net cross-subsidy" basis between CSIRO and ARC-eligible institutions).

Back to Top

In areas such as agent-based modelling, visualisation and data analysis there is a great demand among the application-oriented participants and end users for help with understanding the concepts, methods and tools that are being developed to analyse complex systems. A Data and Visualisation Support Officer will be employed to assist with the diffusion of new techniques and software to the user communities and Tutorial Workshops will be organised to familiarise users with the concepts and train them in the use of new methods. The Summer Schools (see below) will also play a role in this process.

Back to Top

The biennial Bathurst Complex Systems Summer School, organised by Charles Sturt University, and the Canberra International Physics Summer School, organised by the ANU Centre for Complex Systems form a basis on which an annual complex systems summer school series will be built. Not all Physics Summer Schools will have a strong COSNet component, but it is proposed that in early 2005, the World Year of Physics, the Bathurst and Canberra summer schools combine to run a Complex Systems Summer School in Canberra. A similar summer school will be run under the auspices of COSNet in each subsequent year at Bathurst, ANU or at another node.

The first week of each summer school will be devoted to short courses in the basics of complex systems science, presented by Australian and overseas experts. The second week will have more advanced material and hands-on training.

Back to Top

Where appropriate, COSNet may contribute as a sponsor in suitable conferences and workshops organised by other ARC Networks or other groups, and similarly will seek external sponsorship for its activities.

Back to Top

Many of the advances in our understanding of complex systems have come from the matching up of mathematical paradigms with observations in the physical world. Concepts such as chaos and integrable systems, self-similarity and fractals, and self-organisation have been rigorously studied in abstract models, and then applied to physical systems by searching out the appropriate control parameters and observable quantities. This process involves interaction between theorists and experimenters, and a great deal of model building and testing. The data acquisition and analysis for real complex systems, like fluids, plasmas, and granular systems, is a specialist area in itself, involving the use of advanced techniques for time-series and spatial analysis, wavelets, singular-value determination and other transform techniques, statistical database analysis, pattern recognition, and data mining. Many of these techniques have been developed and refined in response to needs in complex systems studies, and have found wider applications in industry and finance.

Progress in complex system science can be significantly accelerated by encouraging and developing iterative theoretical-experimental research projects in the areas where both passive (observations) and active experiments are possible. These areas include turbulent plasmas (eg in the H-1NF Major National Research Facility at ANU), geophysical experiments and observational facilities (eg the rotating water tanks in the Geophysical Fluid Dynamics Laboratory at ANU, the CSIRO's OzFlux micrometeorological observation network, and the SuperDARN auroral radar network). We thus propose to support a specialist in data analysis and visualisation to aid network participants in analysis and modelling of experimental data for comparisons with theory.

Back to Top

Computational modelling is a core feature of complex systems research, and, increasingly, use of high-performance computing is necessary to study complex systems in greater depth. Many COSNet users use the facilities of APAC, and COSNet will assist with networking these users to take maximum advantage of the computational and visualisation resources available and to develop innovative computational strategies and tools.

Back to Top

The network links the key players in general complex systems thinking in Australia with many others specialising in aspects of complexity (see D3), or in application areas where there is a thirst for knowledge of how to grapple with complexity. This network provides the opportunity to value-add the scattered expertise already developed in Australia in specialised areas through an ambitious interaction program organised in a matrix of cross-cutting General Themes:

linking broad Application Areas:

1. Complex Physical Systems

- astrophysics and cosmology

- plasma physics and space science

- fluid physics, industrial applications, weather and climate

- condensed matter physics, hard and soft

2. Complex Biological Systems

- brain dynamics and biophysics

- environmental biology

- genome/phenome

- immune system

3. Complex Computational Systems

- artificial life (simulation of living systems)

- evolutionary computation (new algorithms)

- pattern recognition, datamining, robotics

- DNA Computing

4. Complex Socio-Economic Systems

- human dominated ecosystems

- economics and "econophysics"

- epidemiology and social networks

- infrastructural networks

Back to Top

A very brief summary of the general themes follows. More extended summaries are found on separate pages.

Most systems in nature are in non-equilibrium states and undergo irreversible transformations, for which a description through classical thermostatics and equilibrium statistical mechanics is impossible. Such irreversible processes (for example heat and mass flow in response to temperature and concentration gradients) are describable theoretically only through non-equilibrium statistical mechanics and/or irreversible thermodynamics, subjects which have been well developed in mainstream physics, but which have yet to find their full potential in analysis of complex systems. Of central importance is the entropy production, which according to the second law can never be negative, and which exhibits an extremal nature under certain circumstances. This enables a system, however complex, to be analysed by a single all-encompassing variational principle. Empirical evidence suggests that this approach may provide a useful general explanation of the observed behaviour of a wide range of complex systems, from photosynthesis to climate patterns. If predictions of the behaviour of such complex systems could be made in this way, the practical consequences and theoretical implications would be immense.

Irreversibility is also an essential component in the phenomenon of emergence, another being nonlinearity. Questions related to the reasons behind emergence, and the detection of organisation and of emergence are important in a wide range of fields, including evolutionary dynamics, artificial life, neuroscience, phase separation, self assembly and turbulence.

Back to Top

A comprehensive understanding of fluid turbulence remains one of the grand challenges in science and engineering. This is especially so for non-Newtonian fluids, plasmas, and reactive flows; not only because of their growing economic importance in Australia and world-wide, but because of their intrinsic, highly nonlinear feedbacks. Fundamental problems related to onset, growth, energy spectra, multi-scaling, and structuring of turbulence are actively researched across disciplines such as engineering and environmental fluid dynamics, space and astrophysics, and plasma physics. In mixing and heat and mass transfer applications turbulent transport has long been the big issue, and research in these areas is driven by a need to model, predict, and control accurately both the onset of turbulence and the energy distribution in fully developed turbulence. Dynamical systems theory and high-performance computational techniques have an essential role in this endeavour. Because of its breadth of application the study of turbulent fluid and plasma flows is a truly multi-disciplinary venture in which advances in one area synergise developments in other areas.

Back to Top

Complex systems often evolve and interact on scales spanning several orders of magnitude in space and in time. In recent years, multifractal scaling analysis has proven useful in uncovering scaling properties in a diverse range of temporal and spatial data sets, the statistics of such systems being non-Gaussian. Scaling laws are also a common feature in nonlinear complex systems that evolve self-organised structures — self organisation driven by power law scaling is called Self-Organised Criticality (SOC), the paradigmatic example being the sandpile. Applications of multi-scale analysis range from financial data analysis to diffusion in biological media, earthquake prediction and turbulence in fluids and plasmas, including space physics applications.

Back to Top

Networks are everywhere: neurons, individuals, societies, economies, and environments are all interconnected by intricate and dynamically changing networks. Likewise, vast networks of interacting cells and genes control growth and development. Patterns of network connections determine many features of complex systems. To deal with large networks requires an understanding of how global organisation and behaviour emerges from local interactions. End users need tools to help them plan and influence the behaviours that emerge from networks. An improved understanding of networks and how to manage them would benefit many areas of activity, including economics and commerce, advanced computing, environmental management, biotechnology, medicine and government.

Back to Top

A cellular automaton (CA) models any world in which space can be represented as a uniform grid, time advances by steps, and the "laws" of that world are represented by a uniform set of rules which compute each cell's state from its own previous state and those of its nearby neighbours. Agents are automata that are somewhat more complex in their internal processing and consequently in their behaviour. CA models have been used in many areas of the physical and spatial sciences, biology, mathematics and computer science, as well as in the social sciences. There is no generally agreed definition of what an agent is, the term usually implying an autonomous, intelligent entity that may interact or communicate with other autonomous, intelligent entities. Applications range from artificial intelligence and robotics to modelling the interaction of humans with ecosystems.

Back to Top

Some Big Questions

(You can help contribute to discussion on these questions, or suggest new ones, on COSNet Wiki page: The Big Questions)

How did life emerge from primordial chaos?

One of the big questions facing science is how living arose from nonliving matter. In biology, the question is what were the first replicators, for example RNA, proteins or clay molecules. In astrophysics, questions arise as to what are the physical preconditions for life, so as to ascertain how common life might be in the universe. Physics concerns relate to what boundary conditions on nonequilibrium dissipative processes are necessary to generate complex structures. A theory of entropy production in far from equilibrium processes is an essential ingredient in understanding what is life. A number of COSNet researchers are interested in entropy in far from equilibrium systems, including named Participants Farquhar, Prokopenko, Robson and Dewar. Complex systems theory has examined autocatalytic sets of chemical reactions. There is an error threshold limiting the complexity of these autocatalytic sets, which is perhaps 3­4 orders of magnitude simpler than the simplest living organism. This complexity gap is the challenge facing the complex systems community. Part of the answer will presumably lie in understanding the formation of self-organising complex structures such as lipid membranes. Studies in chemistry and artificial chemistry, such as those done by Duraid Madina will shed light on these aspects. Understanding how evolution can produce complexity in an open-ended fashion will also be another key step, a question currently being studied by named Participant Standish and David Newth. Other named Participants who can contribute to a multidisciplinary attack on this problem range from Freeman on astrophysics to Ball, Bradbury, Farquhar and Wilson with expertise in biology.

Back to Top

The energy released by fusion reactions in the sun not only provides the high-grade radiant energy that drives the self-organising, life-supporting environment of the earth, but also causes more violent events that give rise to “space weather”. There is also a worldwide research and development program for producing power from fusion, a far more benign source than fission. As in the sun, this is complicated by turbulence.
Understanding, predicting and controlling these entropy-producing processes is essential to our long-term future on this planet and requires a multi-disciplinary complex systems approach. One of the biggest problems in fusion is turbulence—how to predict it and control it. Fusion plasmas, space environment, Earth’s atmosphere and oceans, and industrial fluid mechanics are real-world systems whose dynamics is turbulent. Named Participants working on theoretical and experimental aspects of turbulence and its applications are Ball, Chong, Denier, Dewar, Finnigan, Frederiksen, Griffiths, Metcalfe, Raupach, Roberts, Robson, Shats, Soria, Vincent and Vladimirov.

Back to Top

There is a growing realisation (see e.g. Geoff Davies’ new book Economia: New economic systems to empower people and support the living world) that the “business cycle” may be a manifestation of instabilities embedded deeply in the economic system: fairly obviously in financial markets, but also in the monetary system, and that complex systems analysis is required to understand and better control the economy. Ideas for controlling and eliminating these instabilities will need to be tested theoretically within the context of a dynamical system that accounts for strong couplings and feedbacks which generate complex behaviour.
Since 1990, physicists have created a new cross-disciplinary field spanning statistical physics, complexity science and economics: Econophysics, linking scientific academia with the finance sector. Among the named Participants, Aste, Batchelor, Batten, Henry, Heyde, Prokopenko and Standish are working on aspects of this field, as well as others in the Register, in particular Tiziana Di Matteo who is organising the international workshop on econophysics “Bonzenfreies Colloquium II” in 2005.

Back to Top

The greatest challenge facing artificial intelligence is ironically common sense—the joining together of many disparate pieces of information. With the vast increase of archived data and the major threats to national security and the ever present risk of new pandemics, there is an urgent need for such giant intelligence. This can only come from methods of integrating intelligent, adaptive agents on a huge scale. Key mathematical issues are the optimal connectivity, being addressed in a current Discovery Project by Snyder and Bossomaier and the agent integration and evolution mechanisms, the subject of a DP funding request by Snyder, Bossomaier and Mitchell. Named Participants working in problems of artificial intelligence and brain function are Abbass, Bossomaier, Liley, Prokopenko, Wiles, Robinson and Standish.

Back to Top

The nation will benefit through better science across a broad canvas. The problems that the Network will tackle are hard problems, ones that have generally proved intractable with traditional reductionist approaches. Complexity science offers the hope that many of these problems will be better understood, and that some will be solved.

By choosing hard problems and aggressive approaches, the Network aims to achieve major rather than incremental advances in our knowledge base.

The Network is designed to generate a broad national benefit by:

• creating a critical mass of complex systems researchers in Australia
• closely coupling that cadre of scientists with similar leading-edge ventures overseas
• encouraging the spread of complex systems thinking, tools and techniques across a widening application base
• transforming the perception of problems through a complexity world-view
• engaging new stakeholders to use complexity approaches

In particular, the work of the Network will foster greater fundamental understanding, prediction and control of many hitherto intractable problems in important complex dynamical systems and complex adaptive systems. Turbulent flows in oceans, the atmosphere, plasmas, galaxies and industrial processes will be better understood, with the potential for better prediction and, in some cases, control. This will create immediate economic benefits in some cases such as the oceans or climate or industrial processes, and offer long-term benefits in others such as plasmas. The behaviour of brains, societies, economies and ecosystems will be better understood, and intervention strategies could be better designed. Better models of such CASs will lead to benefits in human health, national security, economic well-being and sustainability. New mathematical and algorithmic tools will be developed and rigorously tested on real problems.