Research Training Group 1194 "Self-organizing Sensor-Actuator-Networks"

Research Topics

Das Forschungsprofil des Graduiertenkollegs

Verteilte Netzwerke, bestehend aus hunderten, möglicherweise tausenden von miniaturisierten und autonomen Sensor-Aktor-Systemen werden die Beobachtung und Regelung verteilter Phänomene revolutionieren. Typische Anwendungen sind die Beobachtung großer geographischer Gebiete, intelligente Gebäude, mikroskopisch kleine Sensoren und Aktoren im oder am menschlichen Körper und Sensoren zur Überwachung von Geräten und Maschinen. Im Gegensatz zur Verwendung einiger weniger makrokopische Sensoren wird zum einen eine weitaus höhere Auflösung erzielt. Zum anderen kann durch die große Anzahl von Systemen auch bei einer geringen Zuverlässigkeit und Verfügbarkeit der individuellen Sensor-Aktor-Systeme die notwendige Toleranz gegenüber Fehlern und Ausfällen erreicht werden.

Zur Verdeutlichung wird die Schadstoffanalyse in einem Fließgewässer mit Hilfe eines Sensor-Aktor-Netzwerks betrachtet, Abbildung. Dabei ist die Verwendung eines heterogenen Netzwerks sinnvoll, wobei die Mehrzahl der Sensor-Aktor-Knoten nur über geringe Rechenressourcen verfügt und nur über kurze Reichweiten mit geringer Bandbreite zu ihren nächsten Nachbarn kommunizieren kann. Zusätzlich enthält das Netzwerk einige wenige komplexe Knoten, welche mit einem globalen Positionierungssystem und einer weitreichenden Kommunikationseinrichtung ausgestattet sind. Dieses Beispiel demonstriert einige der wesentlichen Herausforderungen bei der Entwicklung eines Sensor-Aktor-Netzwerks. Dies beginnt mit der Ausbringung der einzelnen Knoten mit dem Ziel einer ausreichenden Abdeckung und Erfassung des interessierenden Phänomens, in diesem Fall des Schadstoffprofils. Offensichtlich ist ein solches Netzwerk durch den Strömungseinfluss starken Änderungen in der Topologie unterworfen, was spezielle Routingmechanismen für die Multi-Hop Kommunikation, d.h. von Knoten zu Knoten, erfordert. Um die oben erwähnte Abdeckung zu erreichen, werden Aktoren, z.B. Propellerantriebe, verwendet. Weiterhin sind für eine Rekonstruktion der gewünschten kontinuierlichen Schadstoffverteilung aus orts-, zeit- und wertediskretisierten Daten eine Lokalisierung aller Sensor-Aktor-Knoten und geeignete Verfahren zur dezentralen Informationsverarbeitung innerhalb des Netzwerks erforderlich.

 Hardware/Software - Systemintegration

Beside communication and information processing there is a third emphasis on hardware/software system integration based on reconfigurable hardware circuits. This form of adaptive hardware is very flexible, of high performance and additionally provides positive features like energy efficiency and reliability which is of central importance for sensor/actor networks. Further requirements consist of realtime capabilities and self organization. In this context, especially parameters from main area I – Information Processing – have to be considered in order to be able to perform an adaptation of the hardware/software system depending on the situation in realtime. Like in the other two main areas the third main area is divided into sub projects which focus on investigation of unresolved problems in the area of hardware/software system integration for sensor/actor networks. Sub project H1 investigates application specific hardware architectures with the goal of development of energy autarchic sensor/actor nodes, whereas focus is on heterogeneous node architectures. A further emphasis in sub project H1 is on the development of systems for energy harvesting and storing. Questions concerning self-organization of power loss minimized reconfigurable hardware which can be used within the nodes of a network are considered in sub project H2. Finally, sub project H3 is the bridge between the areas Communication, Information Processing and Hardware/Software System Integration. Here, middleware architectures with regard of self-organization, self-optimization and self-healing are explored while ensuring realtime characteristics.

 Information Processing
 There are two distinct tasks in the area of information processing within sensor-actor-networks. These are pure measurement tasks on the one hand and tasks which require coordinated action from all participants within the network on the other.

The query-plan from subproject K4 is transferred by the self organizing middleware of subproject H3 to the individual sensor-actor-nodes via the content based addressing and transport system of subproject K3. These nodes acquire the requested data which creates a massive amount of distributed information that can only be saved within the network for a short amount of time. Furthermore, it cannot all be transported through the network in accordance with subproject K1's findings. Thus the data must be processed “In-Network” to compress and filter the information. Subproject I3 will be tasked with finding “In-Network” algorithms which find an ideal balance between communication and information processing so as to use the available energy with maximum efficiency. This will require close interaction with subproject K1.

Often times information will be requested which can not be directly acquired. The methods for constructing such information from network measurements will be the task of subproject I1. To this end decentralized, model based methods with integrated uncertainty considerations are to be examined and developed by the I1 subproject. The localisation methods developed within subproject K1 will be needed for this.

In the case of actor manipulation tasks, a given job is distributed to relevant sensor-actor-nodes which then cooperatively find a solution in a self-organised and decentralised manner with regard to potential real time constraints. This facet is handled by subproject I2 which will use specialized swarm techniques in order to control the high level of complexity and interaction. This also heavily requires the distributed models of subproject I1.
 Communication
 Communication within a distributed sensor-actor network inherently falls under a different set of constraints than conventional communication networks such as the cell network or cable based networks. As the majority of communication takes place wirelessly, energy consumption in relation to bandwidth must be heavily considered while developing communication protocols. Furthermore, communication is heavily influenced by how central information processing is handled.

The K2 subproject will concerne itself with fundamental communication aspects such as communication capacity, protocols and architecture. Quality assurance aspects such as timeliness and reliability will also be discussed. Subproject K2 will be dedicated to questions of topology and localisation. Due to quick changes caused by mobility, energy saving modes etc. this field represents a significant challenge. The primary focus will be on algorithmic aspects which should result in close interaction with subgroup K2. Also close connections to area H concerning reprogrammability are expected.

Subprojects K1 and K2 form the basis of subproject K3 which is much more concerned with application specific aspects such as content sensitive addressing. Furthermore, aspects of robustness and privacy will be discussed which are heavily interdependant with areas I and H.

Based on discoveries made in subproject K3, subproject K4 will analyze how declaritive requests are processed within sensor actor networks. This requires close connections to K2, as it relies heavily on the topology of the network. Also there are close ties to subprojects I1, I2 in respect to information processing.