Democratic processing: Mastering the complexity of communicating systems
Hylke W. van Dijk · University of Twente Research Information · 2004
ApproachOur approach is based on a three notions.First, a system has many beyond our reach properties, which are valued differently by different stakeholders.We adopt an "ontology of the world" [Bunge, 1977] to put in perspective different views and different predicates of the system.Second, we conjecture that a complex system of communicating entities should be addressed as an organism with distributed coordination.Rather than applying a hierarchical organisation we suggest a heterarchic organisation of the coordination, where the dominance is demand driven.Components in the system therefore should be able to abstract their distortion, capacity, and resource utilisation metrics, should be able to adapt to changes in their milieu, and should be willing to cooperate with components in their immediate environment; all components should, like the system, be truly context aware.The above arguments are merged into a framework (ARC) for coordinating a communicating system.Third, the compositional property is an important aspect for the adequate development of any complex system.However, context-aware components are necessarily indeterminate and consequently lack the compositional property.In order to retain this compositional property we develop a model of computation (CAPN) that concentrates the context-dependency of an entity in a control stream.The autonomous processes in this model are flexible in the sense that they can asynchronously adapt to and cooperate with their milieu.We coined this type of operation "democratic processing". ResultsThe study of developing complex communicating systems is a multi-something problem, where something can be anything, ranging from disciplinary development to objective optimisation or from input to output.We succeeded in structuring the multisomething problem in three themes and provide practical approaches and case studies for each of them.The three themes are communication, coordination, and composition.For communication we rely on a multi-view representation of the system.Each view targeting specific concerns and conveying dedicated information.For coordination we developed the Adaptive Resource Contracts (ARC) framework that supports flexible designs with non-functional aspects made explicit: distortion, capacity, and resource utilisation.ARC component are truly context-aware; they implement abstraction, adaption, and cooperation.For the compositional property, ARC isolates the indeterminate behaviour of a component through a so-called oracle.Because of this isolation, the network retains its compositional property.Our Context-Aware Process Network (CAPN) model is an explicit model of computation that implements ARC concepts for the class of streambased applications with occasional event handling.Beuningen, 1849Beuningen, ,1958]].Picture a visitor wearing the Ubicom terminal: a small device connected to a see-through display.The device has a wireless interface to the hidden backbone of the Ubicom system: ubiquitous computation and storage resources in the backbone network infrastructure.The system is context aware: it knows the position and orientation of the user as well as the user's preferences.The Ubicom system situates virtual graphical objects in overlay with real-world artefacts by projecting them on the see-through display such that, from the point of view of the user, the objects seem to be connected to real-world objects: visual augmented reality.The Ubicom system executes a tour guide application.A small virtual creature acts as a guide who suggests possible interesting pieces of art.He leads the way to the old masters section.Here, the visitor enjoys a painting called "De kwakzalver (The Quack)" by Dou [1652].The painting is an exemplar of the so-called Dutch genre paintings, which are full of symbolic references.When the visitor gazes at the painting, vivid animations clarify the scene pointing out the symbols, which refer to the contrast between luxury and austerity, or indicate dim-wittedness and deception.The visitor can alos choose to study the various hidden sketches which were the basis of the painting as it is today.These so-called underdrawings can otherwise only be seen separately using infra-red, ultrasound, or roentgenographic technology.The new wing of the museum (author's imagination) accommodates the D.I.Y. hall.Here visitors are invited to sculpture and re-sculpture virtual objects of art.The room is equipped with empty frames mounted to the wall and socles placed on the floor.With digital paint and digital clay visitors make abstract works of art.The sculptures are left behind for others to enjoy or to modify.Particular successful collective works of art are conserved and put on display.The leitmotif of this dissertation is the development of a Ubicom system as a system.The development does not take place in isolation; it is part of the Ubicom project, which in turn is part of the aforementioned Ubicom research programme.The Ubicom project addresses the technological issues of the research programme. Complex systemsThe development of a communication system in a multidisciplinary setting such as the Ubicom project is inherently difficult; it is a complex system in a complex setting.The complexity of any system is increased by the interplay of a number of aspects.The following are a few outstanding ones, which when used in combination define a complex system.Scale Large-scale structures tend to obstruct a clear view. Interaction Interactions among components may complicate their coordination. Diversity The involvement of diverse technical domains (disciplines) induces heterogeneity in modelling and research approaches.Irregularity Regular and homogeneous structures or interactions can be mastered by good bookkeeping alone.Irregularity and inhomogeneity complicate systems. FluxThe immediate environment of a system demands that the system is in a constant state of flux. Abstract communicationsThe observation that different views use different definitions but are phrased in similar terms is important.It is equally important to recognise that each view exists for a purpose.The mere recognition of the underpinning problem is the first step of solving it.We use Bunge's systemic philosophical position and corresponding ontological framework [Bunge, 1977[Bunge, , 1979] ] to interrelate multiple coexisting views on a system.Each Hierarchical (authoritative) design methods usually lack adequate support for most of the above-mentioned aspects.One of the problems is their centralised coordination.In order to master the complexity, hierarchical design methods abstract subsystems by modelling their behaviour and performance in a fixed system-wide structure.The structure and models are imposed by a central authority: the architecture team.In effect the whole system is captured in a single (meta) framework.We object to this style of design as articulated in Postulate 1.1.New developments, research results, or new products put on the market will fit in as long as they do not trigger radical changes of the system-wide model.The impact of changing the system-wide model can be quite significant.Notorious are changes of the functional interface for a specific component.