border shadow

The Event Collector Concept (2001)

A distributed Infrastructure for Event Generation & Dissemination in Ad Hoc Networks. Diploma thesis in electrical engineering attended by Prof. Friedemann Mattern, Oliver Kasten and Michael Rohs, fall of 2000.

Abstract

In mobile ad-hoc networks, many applications need to comprehend the environment made up of the participating devices. This thesis designs and prototypes an architecture for distributed gathering and dissemination of network information like link status or node characteristics to extract neighbourship information. Knowledge is acquired by evaluating events generated upon nodes entering or leaving each others radio range. This concept is tested both in a simulation environment as well as on a small hardware platform built around a bluetooth transceiver.

Above the BT-node. To prove the concept this small platform was built with an Atmel 8-bit micro controller (black square in the middle) and an Ericsson Bluetooth module (golden package on the left). Additional features are 4 LED for signaling, a serial connector and sensor I/O capability. The green thing on the far left is the antenna. The battery is hidden underneath.

Introduction

The shift into the 21'st century was accompanied by a shift of paradigms in information technology. The personal computer age which thrived in the 1990's led over into the age of ubiquitous computing. The omnipresence of information technology is revolutionizing usage of everyday-things. Appliances are built with more and more intelligence. In a next step, these devices will begin mutual communication over wireless links on an ad-hoc basis. Undreamed-of applications will revolutionize daily living in the age of pervasive computing.

 

Wireless ad-hoc networks differ fundamentally from their wired counterparts. Mobile wireless ad-hoc networks (MANETs) do not rely on a base station or central control. Most of the components are mobile, thus network topology changes constantly. As nodes join and leave networks frequently, predictions about the connectivity between any two nodes at a given time is a difficult task. Moreover, connections between devices are unstable, not secure and prone to errors.

Communication takes place in a peer-to-peer fashion and interaction must be enabled without prior configuration. Nodes deployed in places without existing infrastructure must cooperate in a common exploration process and keep on doing so in the ever changing environment.

Nodes which make up ad-hoc networks may be of different kinds. Anything from supercomputers down to credit-card based low-power devices may want to interact.

Since joining nodes are a priori unknown, there is little or no information about the node characteristics or capabilities of partnering devices.

 

In this thesis, we propose a distributed architecture where nodes cooperate to collect and distribute information about their environment. Situation awareness is a fundamental prerequisite for further services such as routing, tracking or navigation.

 

Logical Structure

The EventCollector concept consists of four layers:

  • In the "Discovery layer", information about a node's vicinity is collected. Various types of properties like changing link states, sensor data or information about the existence of nodes which are located in radio distance may be gathered. Generated data is encapsulated into events which are passed up one layer for storage and propagation. In our implementation each Bluetooth node perjodically makes an inquiry to search for neighbouring devices.
  • The layer responsible for "knowledge sharing" manages local storage of events and handles propagation of event data. To share collected network information, events are exchanged between adjacent network nodes and thus disseminated throughout the network. The basic idea is: if everyone shares its knowledge with others, at the end, everybody knows more. The lower two layers form an entity called EventCollector.
  • The "Information Processing" layer processes collected events provided by the EventCollector and acts as server to entities in the "Application Layer". Here some algorithm tries to extract useful information out of the collected data. For example it tries to find out how long a neighbour already has been there or how often it appeares and disappeares again.
  • Applications may exploit this knowledge to their benefit. For example, the history of connections can be used in multi-hop routing algorithms to determine a route between network nodes. The same information may be used to compute approximate location of mobile nodes, if the network comprises a number of stationary nodes with known location (e.g. printers, set-top boxes, or specialized "responder beacons" nodes).

Physical Structure

Such an EventCollector architecture was implemented for this thesis. Events are generated and disseminated by EventCollectors in either a simulated or a live environment. In the live environment, the mobile nodes use Bluetooth[1] as communication medium. Gateways make event information accessible to applications running on PC's, where further processing is done. As an example, a topology snapshot and measures for mobility and average connection time is extracted out of the network information. This information is provided to an application for graphical display.

Architecural overview: On the left the live environment with the Bluetooth nodes, on the right the Java simulation environment (BTSim) and in the middle the extraction algorithm (Vicinity) and display application (Connection Graph). The Vicinity consumes events generated eather in the live or in the simulation environment. For a larger view click on the image.