Let’s start by defining the Perplexus platform as a network of ubidules, i.e., ubiquitous computing modules. Such a computing platform is the main technological objective of the project and should enable the simulation of large-scale complex systems and the study of emergent complex behaviors in a virtually unbounded wireless network of computing modules. At the heart of these computing modules we will use a custom reconfigurable electronic device capable of implementing bio-inspired mechanisms. This reconfigurable circuit will be associated with rich sensory elements and wireless communication capabilities.

- The wireless link between the ubiquitous computing modules (ubidules) will guarantee a scalable solution to the complex systems simulation platform. The number of ubidules, their computing resources, and the way they are interconnected will be adapted to the particular needs of the modeling application. The wireless nature of the communication between the modules will enable a virtually unbounded and scalable computing platform. In the case of simulating biologically plausible large-scale neural models this will constitute a clear advantage over classical emulation or simulation alternatives, since the dense connectivity pattern among individual units constitutes the major factor preventing their analysis with actual real-time stimuli. Furthermore, wireless communication will implicitly guarantee the notion of locality of interactions between all entities constituting the platform to be simulated. Considering a software simulation, this parameter has to be explicitly modeled which in turn leads to an additional factor impacting scalability.

- The large capacity of computing power and flexibility will facilitate the simulation of classical complex system models (discrete grid-world models or differential-equation-based models), but at the same time the inherent bio-inspired capacities of the computing modules will make it an ideal framework for simulating biological systems. The extendable architecture of the platform will enable the simulation of large-scale complex systems based on “perplex worlds”, extendable and virtually-unbounded discrete models resulting of this new simulation and modeling platform.

- The platform will have self-adaptive features both at the functional and at the sensor level. The self-plasticity at the functional level is required to permit the construction of actual distributed systems that may adapt to the environment without the need for a centralized and external control unit. On the other hand, the sensor plasticity will allow the sharing of the sensors actually present in the system irrespective of their physical location.

- The wireless network of ubidules will be able to communicate with neighbor ubidules and share/distribute/delegate tasks.

- The wireless network of ubidules can be seen as a population of interacting individual organisms that can exhibit emergent complex behaviors. Moreover, groups of ubidules might specialize through self-organization and evolutionary mechanisms to form independent “species” of computing devices, or complex organisms treating a common task, formed by a set of ubidules, where the limits of the structure of the organism is determined by the self-organizing process itself.

We can distinguish two approaches to tackle with complex systems. On the one hand, the top-down approach which functions in a divide-and-conquer manner. A problem is divided into sub-problems that are individually considered. Such an approach has the problem of missing the fact that complex behaviors need not have complex roots, and models are not scalable. On the other hand, bottom-up approaches do consider explanations of complex behavior via simple, local component interactions in a decentralized manner, generally providing a scalable model of the complex system. In this project, we will consider challenging modeling problems following a bottom-up approach. The modeling framework we will develop in this project will be a breakthrough in the domain of systems engineering, given that it will enable an unprecedented platform for modeling complex systems.

As far as the computer models of complex systems are concerned, we can also distinguish two approaches. On the one hand, there are continuous models based on differential equations. These models generally suppose a high abstraction level and do not integrate a spatial dimension. On the other hand, there are discrete models that introduce individual entities called “agents” or “particles”. Agent-based modeling is very often based on discrete “grid-world” models. In this project we will explore the use of “perplex-world” models, virtually unbounded discrete models realized by interconnecting many ubiquitous computing modules (ubidules) and by interacting with real-world sources. We will determine the range of applications and the advantage of such a framework for simulating complex systems in the domains of realistic neural models, the dynamic emergence of culture, and collective robotics. In particular, we have identified challenging modeling issues like a) the use of “dynamical modeling spaces” resulting of the dynamical topology of the modeling infrastructure, b) the use of unbiased variable values coming from the environment via the sensors of the ubidules, and c) the use of unexpected new information or loss of information resulting from ubidules dynamically joining/leaving a given modeling framework.

A second major scientific objective is to study the emergent phenomena arising from the strong interaction between our hardware infrastructure and the real environment. The PERPLEXUS platform can be seen as a complex system itself: it is a population of artificial organisms interacting with the real environment. We will study emergent organization mechanisms that will enable the formation of two kinds of structures: a single distributed complex organism or a population of collaborating-competing individual organisms.

A third scientific objective is the study of the emergence of an “artificial culture”. The ubidule ubiquitous computing modules learn by interacting with the environment and with the user. This learning process might lead to the acquisition of “concepts” that might be communicated to neighbor ubidules and diffused through the perplex platform. The sharing of such concepts might prosper and reach a stable state throughout the platform. Such a stable state or concept consensus might be seen as an “artificial culture” shared by the ubidule modules of the Perplexus platform.

We have identified three domains where our modeling infrastructure will prove its usefulness as a powerful and innovative simulation tool:

1. Computational neuro-genetic modeling

There is experimental evidence that the cerebral cortex develops as a whole rather than regionally. Genetic programs are assumed to drive the primordial pattern of neuronal connectivity through the actions of a limited set of trophic factors and guidance cues, initially forming excessive branches and synapses, distributed somewhat diffusely. The embryonic nervous system is refined over the course of development as a result of the twin processes of cell death and selective synaptic pruning. The cutting-edge of this modeling application lies in the possibility to simulate several small-scale networks, each one following its faith in terms of synaptic pruning and changes of synaptic strength during a certain amount of time, that are given the possibility to interact and influence dynamically each-other information processing in order to let emerge new properties. This goal should be achieved by means of our new bio-inspired device, that allows the embedding of neural networks with spiking neurons and modifiable synapses.

2. A Connectionist, embodied and situated agent-based approach for studying the dissemination of culture

The prevailing approach to study the dissemination of culture has relied on a view of an unbounded and unbiased mind, equally open to any kind of cultural content. The objective of this modeling application is to challenge the prevailing approach by using an agent-based approach coupled to artificial neural networks representing “cognitive biases” in a systematic framework. We will take advantage of the Perplexus pervasive computing infrastructure by implementing complex models of culture dynamics out of dynamically interacting computing modules and by exploiting signals coming from the environment via the sensors of the pervasive units.

3. Social robotics

Cooperative robots are constantly interacting not only with the dynamic environment, but also naturally with each other and with the persons who are using them. It seems obvious that the number of distributed autonomous robotic system applications will increase rapidly as the technology and knowledge improve. Within few years, various robot societies will move out from research laboratories into everyday life. Normal applications will include tasks like cleaning, monitoring, delivering, etc. Besides that, a number of more revolutionary applications will probably also emerge. The objective of the social robotics modeling application is to study emergent behaviors resulting from the multiple interactions among robots and with the environment. In particular, the the emergence of the robot's adaptability to the human society.