Int. J. Autonomous and Adaptive Communications Systems, Vol. 4, No. 1, 2011 39
Copyright © 2011 Inderscience Enterprises Ltd.
Emergent engineering: a radical paradigm shift
Mihaela Ulieru*
Adaptive Risk Management (ARM) Laboratory,
Faculty of Computer Science,
University of New Brunswick,
P.O. Box 4400,
Fredericton, NB E3B 5A3, Canada
E-mail: ulieru@unb.ca
*Corresponding author
René Doursat
Institut des Systèmes Complexes,
CNRS and CREA,
Ecole Polytechnique,
57–59, rue Lhomond,
Paris 75005, France
E-mail: rene.doursat@polytechnique.edu
Abstract: We shed light on the disruptive advances brought by the ubiquity of
computing and communication environments, which link devices and people in
unprecedented ways into a new kind of techno–social systems and
infrastructures recently named ‘cyber-physical ecosystems’ (CPE). While
pointing to fundamental biases that prevent the traditional engineering school
of thought from coping with the magnitude in scale and complexity of these
new technological developments, we attempt to lay out the foundation for a
new way of thinking about systems design, referred to as emergent
engineering. One major characteristic of CPE is that, given their very nature,
they cannot be a priori defined but rather emerge from the interactions among a
myriad of elementary components. We show how this emergence can be guided
by balancing positive and negative feedback, which tunes the growth of new
configurations and adapts the system to sharp and unexpected changes. Rather
than attempting to design the system as a whole, the components of the system
are endowed with capabilities of dynamic self-assembly, disassembly and
re-assembly to enable ‘evolve-ability’. As paradoxical as it may seem to the
classically trained systems engineer, this new attitude of the designer as an
‘enabler’ (vs. ‘dictator’ of a system’s blueprint) allows the system to seamlessly
adapt its development and evolve to meet dynamic goals and unexpected
situations in an anticipative manner – an impossible feat under the traditional
approach. To the extent that it produces new functionality, the proposed
method enables a system to evolve via its ability of pervasive adaptation.
Emergent engineering lies at a boundary where theoretical discovery meets
systems engineering, computing and communications into a new convergent
science of complex systems design. It currently transforms systems and
software engineering by embracing various highly interdisciplinary
perspectives.
Keywords: complex systems; emergent engineering; emergence and
self-organisation; CPE; cyber-physical ecosystems; developmental biology;
co-evolution.
40 M. Ulieru and R. Doursat
Reference to this paper should be made as follows: Ulieru, M. and Doursat, R.
(2011) ‘Emergent engineering: a radical paradigm shift’, Int. J. Autonomous
and Adaptive Communications Systems, Vol. 4, No. 1, pp.39–60.
Biographical notes: Mihaela Ulieru is the Canada Research Chair in Adaptive
Information Infrastructures for the eSociety and Director of the Adaptive Risk
Management Lab, conducting research in complex networks as control
paradigm for complex systems at the University of New Brunswick. She
currently champions the area of emergent engineering and its application to
emergency response management and networked enabled operations. She is the
founder and leader of several international research consortia, and was
appointed on several national and international advisory boards and review
panels, among which the Scientific Council of the EU, NCE Intelligent
Manufacturing (I*PROMS), the EU FP7 Proactive Initiative on Pervasive
Adaptation (PERADA), Australia’s Digital Ecosystems and Business
Intelligence Institute, and Singapore A*STAR. In 2007, she was appointed by
the Minister of Industry as a member of the Government of Canada’s Science
Technology and Innovation Council.
René Doursat is Director of the Complex Systems Institute, Paris Ile-de-France
(ISC-PIF) and Full Member of CREA, the research centre in cognitive science
and self-organisation at the Ecole Polytechnique, Paris. Previously, he was a
Visiting Assistant Professor in computer science at the University of Nevada,
Reno. An alumnus of the Ecole Normale Supérieure, Paris, he came back to
academia full-time in 2004, after a segue through San Francisco Bay Area’s
software industry. His research activities address the computational modelling
and simulation of swarm multi-agent systems aimed at a new form of
engineering inspired by biological and social complexity – in particular the
emergence, dynamics and evolution of heterogeneous architectures. He is the
Principal Organiser of the French Complex Systems Summer School in Paris
and several international workshops in complex systems science and
engineering. He is an Associate Editor of IEEE Transactions on Neural
Networks, and an expert reviewer or advisor for several journals, grant
agencies, award juries and curriculum committees.
1 Introduction
Information and communication technologies (ICT) pervading everyday objects and
infrastructures, the future ‘Internet of Things’ (ITU Internet Reports, 2005) is envisioned
to undergo a radical transformation from today’s mere communication highway into a
vast hybrid network seamlessly integrating physical, mobile and static systems to power,
control or operate virtually any device, appliance or system/infrastructure. Manipulating
the physical world will occur locally, but control and observability will be enabled safely
and securely across an overlay network that we broadly refer to as an ‘eNetwork’. Such
eNetworks will enable the spontaneous creation of collaborative societies of otherwise
separate artefacts, referred to as ‘cyber-physical ecosystems’ (CPE).
1
Their examples
range from self-reconfiguring manufacturing plants (Ulieru, 2004) and self-stabilising
energy grids to self-deploying emergency taskforces, all relying on a myriad of mobile
devices, software agents and human users that would build their own eNetwork on the
sole basis of local rules and peer-to-peer communication (Dressler, 2007). In such
‘opportunistic ecosystems’ (herewith referred to as eNetworked CPE) that will make the
Emergent engineering: a radical paradigm shift 41
Internet of Things, distributed systems at various levels of resolution, ranging from single
devices to spaces, departments and enterprises, are brought together into a larger and
more complex ‘system of systems’, in which the individual properties or attributes of
single systems are dynamically combined to achieve an emergent desired behaviour of
the synergetic ecosystem.
The dramatic progress of CPE technologies is envisioned to reach unanticipated
levels of complexity, beyond the boundaries of the disciplines that conceived their
components (CPS, 2008). This challenges the traditional engineering school of thought in
disruptive ways, given that, by their very nature, CPE cannot be a priori defined, but
rather emerge from the interactions between individual systems’ (and people’s),
interactions facilitated by the eNetworks. This requires to drastically revise the traditional
top–down perspective on system design and control, which aimed at imposing order
exogenously, telling each element of the system what to do at every step through
predetermined strategies, and assuming that all possible situations the system might
confront are knowledgeable in advance. Instead of fighting it, eNetworked CPE could be
managed by ‘riding the wave’ of their own complexity and rather let systems grow,
function and stabilise – even adapt and improve – endogenously, in a ‘bottom–up’
fashion.
2 Towards a new way of thinking about systems design
We address the radical shift of paradigm in systems and software engineering caused by
the irruption of ubiquitous computing and communication environments. The accelerated
expansion of eNetworks, tightly linking systems and people in unprecedented ways, has
enabled a spontaneous and uncontrolled ‘bottom–up’ emergence of hyper-distributed
CPE. Machines, critical infrastructures, softwares and users are now blended at a
magnitude and level of complexity that exceeds the traditional ‘top–down’ engineering
mindset. This has puzzled systems and software engineers for some time now and started
a worldwide revolution (IT Revolutions, 2008) that aims at a new way of thinking about
such complex systems. The new quest is to find appropriate methods to manage the
magnitude of scale and complexity of large CPE.
One major characteristic of large interdependent CPE is that, by their very nature,
they cannot be a priori defined but rather emerge from the interactions between
individual machines and people, facilitated by eNetworked communication. Recent
attempts to understand and handle these new types of networks point to an alternative
school of thought in systems and software engineering, questioning the main stream in
disruptive ways. Instead of defining the system and its performance requirements in
advance, following a top–down hierarchical thinking (Figure 1(a), inspired by Carreras
et al., 2009), the engineer must rather act as a facilitator to support and guide the
complex system through its process of ‘self-design’, which generates organisational
structure from the bottom–up interactions among a myriad of elementary components
(Figure 1(b)). As paradoxical as it may seem to the classically trained systems engineer,
this new attitude of the engineer as enabler (vs. ‘dictator’ of a system’s blueprint) allows
the system to seamlessly adapt its development and evolve to meet dynamic goals and
unexpected situations in an anticipative manner – an impossible feat under the traditional
approach.
42 M. Ulieru and R. Doursat
Figure 1 The radical shift in design paradigm: (a) top–down design and (b) bottom–up ‘design by
emergence’ (see online version for colours)
Building on these trends, time is ripe to capitalise on the recent advances in systems
engineering, computing and communications, and develop a new, convergent science of
complex systems design. The significant difficulty of this pursuit is that it lies at the
junction between multiple disciplines: engineering (dynamical systems and control),
communications (networks), computer science (agent-based modelling and simulation
(ABMS)), physics (statistical mechanics) and biology (self-organisation in
morphogenesis, homeostasis and evolution). We need to continue building upon the latest
paradigms, through which the new school of thought is currently transforming systems
and software engineering, towards a global approach embracing various perspectives
from all the above disciplines. We propose to call this unified theoretical effort emergent
engineering.
One major mandate of the new school of thought is to formulate and define the
concepts of emergent engineering from this radically new, interdisciplinary perspective,
as suggested in Lee (2007):
“Today’s computing and networking technologies, however, may have
properties that unnecessarily impede progress towards these
applications}Many of these applications may not be achievable without
substantial changes in the core abstractions}To realize the full potential of
Cyber-Physical Systems, we will have to rebuild computing and networking
abstractions. These abstractions will have to embrace physical dynamics and
computation in a unified way”.
This new school of thought encompasses trends in computing and communications as
well as networks. In this paper, we attempt to lay out the basis for new concepts and
abstractions able to contribute to the development of emergent engineering. Using the
paradigms of complexity science, we rephrase the classical concepts of engineering
design and systems control respectively, in terms of developmental emergence,
adaptation and evolvability found in natural systems to propose a breakthrough approach
to the architecting and control of future eNetworked CPE. We proceed by identifying and
Emergent engineering: a radical paradigm shift 43
responding to several fundamental biases of traditional engineering in Section 3, and
illustrate these new abstractions on a model of self-made network that we propose in
Section 4.
3 Fundamental biases carried on from the traditional engineering school
3.1 Traditional engineering requires a system to be well defined
Generally, engineering is about the design of bounded, static systems that can be clearly
and completely defined around specific operating points or regions. As systems that
continuously adapt and evolve in spontaneous, uncontrollable dynamics, eNetworked
CPE cannot be predefined by the designer, be well defined itself. What characterises such
large-scale complex systems with unpredictable dynamics is that non-trivial, large-scale
order can be produced by simple processes involving interactions operating locally on
simple agents or components. For such systems – termed emergent holarchies in Ulieru
(2004) – ‘becoming’ is ‘being’ (Minai et al., 2006). This stands in sharp contrast to the
classical paradigm in engineering with its clear distinction between the design and
production phase, on the one hand, and the functional phase, on the other hand. Even
systems usually considered to be ‘adaptive’ (such as adaptive controllers or neural
networks) follow this two-phase paradigm, allowing adaptation only in the superficial
sense of parameter adjustment – whereas complex systems change not only their
parameters but also their fundamental structures and processes. This is the essence of the
paradigm shift followed by the new school of thought, and the motivation of our work.
As stated in Carreras et al. (2007), Lee (2007), and Alderson and Doyle (2009), we need
to design for emergence, that is, for systems that fundamentally and continually adapt and
evolve.
As both a system and an evolving concept at the same time, ‘evolution’ for
eNetworked CPE should not only be construed as a method to optimise the system but
more importantly as an intrinsic property of the system to be designed (Carreras et al.,
2007). Most of complex systems engineering research has focused so far on specific
domains such as multi-agent systems (Ulieru, 2004), collective robotics and swarms
(Gross et al., 2006), and networks (Newman, 2006). However, clues towards a general
strategy come from the latest insights into developmental biology (Kauffman, 2008),
where evolution’s profound success is supported by the meta-attribute of evolvability as
the ability of the configuration space (in this case, the space of genotypes or phenotypes)
to produce an endless supply of viable configurations with remarkably few obvious dead
ends. Emergent engineering promotes ‘evolve-ability’ (as per Carreras et al., 2007) as a
new paradigm for designing systems capable of evolving towards dynamically changing
goals by continuously adapting to unexpected situations without human intervention
(Marzano and Aarts, 2003).
Another fundamental insight provided by emergent engineering is that highly complex
functional systems
2
can only arise through evolutionary processes of selection in the
context of actual tasks. This fundamentally contrasts with ongoing efforts to design large
real-time response systems through specification followed by implementation, which is
still the case of even today’s distributed systems, applications and techniques involved in
multi-agent systems, service-oriented architectures or Web 2.0 and semantic Web – a
lingering problem that, for example, ‘organic computing’ is also trying to address
