EDIM: Difference between revisions

From
Jump to navigation Jump to search
No edit summary
Line 99: Line 99:


See also
See also
[http://sar/research/publications/SAR-PR-2006-01/SAR-PR-2006-01_.pdf SAR-PR-2006-01 Development of a Software Distribution Platform for the Berlin Roof Net (Diplomarbeit / Masters Thesis). Bernhard Wiedemann. 73 pages.]
[http://sar/research/publications/SAR-PR-2006-01/SAR-PR-2006-01_.pdf SAR-PR-2006-01 Development of a Software Distribution Platform for the Berlin Roof Net (Diplomarbeit / Masters Thesis). Bernhard Wiedemann. 73 pages.] or [http://www.roonk.de Roonk]

Revision as of 23:18, 30 August 2008

EDIM - An Earthquake Disaster Information System for Marmara

Objective of the EDIM project is to develop an earthquake disaster information system for the Marmara region in Turkey. The already existing Istanbul Early Warning System is extended in terms of area, methodology and data transmission. Novel technologies in seismology, structural engineering, and information technology are developed and tested. The project is funded by the German Federal Ministry of Education and Research.

Partners: GFZ Potsdam, Universität Karlsruhe, KOERI, CEDIM
Roadmap: The project started April 2007
Hardware: Wireless-Devices:WRAP-Board
Software: EDIM:Software

Research Opportunities

This is a list of possible topics for scientific work ("Dissertationen", "Diplom"- or "Studienarbeiten"). Some tasks are also more engineering like and more suitable for "Studienarbeiten". This list is to be extended and will always be work in progress. It ranges from more or less brain-storming to more concrete ideas:

Multi-Channel Protocols for Wireless Mesh Networks

The SoftMAC project allows the implementation of different medium access algorithms using standard Atheros hardware. Goal of the project is to test various proposed algorithms in a real testbed, and develop new ones.

Links & Literature:
SoftMAC Project Homepage
Wu: A Survey of MAC Protocols in Ad Hoc Networks

Time Synchronization in Multi-hop Networks

Geo monitoring applications often require knowledge about the exact time (e.g. UTC) a measurement is recorded. Goal of the project is to evaluate current approaches to external time synchronisation in distributed multi-hop networks - in contrast to internal time synchronization, where the nodes only try to synchronize all clocks within the network. Aim is to achieve high precision (10 msec max deviation) combined with low overhead, using standard 802.11 hardware. In a first step only pair-wise synchronization will be studied, adding network wide synchronization later on. The example application of the study is the SAFER project, where some of the GPS receivers will have no connection, and need to be synchronized over the network.

Possible Research Questions:

  • Do we achieve higher precision when pushing the timestamp functionality down the network stack (e.g. by embedding it in the mac layer)?
  • Can we profit from the high precision timer & synchronization in the 802.11 hardware?

Links & Literature:
http://www.ais.fraunhofer.de/~mock/Sensornetzwerke/DSNetzwerke_clock_12.pdf
Holger Karl, Andreas Willig, Protocols and Architectures for Wireless Sensor Networks, Wiley & Sons Inc. 2005
Blum, Thiele: Precise and Low-Jitter Wireless Time Synchronization (2003)
Römer, Blum, Meier: Time synchronization and calibration in wireless sensor networks (2005)
Yoon, Veerarittiphan, Sichitiu, Tiny-sync: Tight time synchronization for wireless sensor networks (2007)

Coding & Aggregation for MiniSEED Data Streams

In the EDIM and SAFER projects, the raw data of each sensor is stored on the node for a limited time. Scientists should be able to access these raw data streams over the network. Aim of this project is to use coding techniques to reduce the network overhead. First coding techniques for single streams are to be evaluated. In a second step cases where several MiniSEED streams are transferred simultaneously are handled.

Links & Literature:
Network Coding Homepage
He et al: AIDA: Adaptive Application Independent Data Aggregation in Wireless Sensor Networks (2003)

TCP in Wireless Multi-Hop Networks

It is a known fact that TCP does not perform well in wireless multi-hop networks. This is in parts due to TCP's congestion-control algorithm, which reduces the transmission rate upon packet losses. A further problem seem to be inefficient retransmission control and considerable overhead. Still, TCP is the standard transport layer protocol, and many applications are based on it, including the MiniSEED software used in EDIM.

In this project a transparent in-network replacement of TCP has to be developed. Possible ideas are the use of NACK's (negative acknowledgement) instead of ACKs, the usage of recovery servers that buffer lost packets on the path

Links & Literature:
Balakrishnan et al: A comparison of mechanisms for improving TCP performance over wireless links
Stann, Heidemann: RMST - Reliable Data Transport in Sensor Networks

EDIM Routing Protocol

Hundreds of different routing protocols were proposed in literature. However, most of them have never been actually used in the real world. Goal of this project is to evaluate the different protocols and select one or possibly a few in a structured way for EDIM.

Distributed Backup over Wifi

The Wireless Nodes do also take care of longterm storage (5 to 7 days) of seismological data. Due to the nature of earth quakes, one or several nodes could be destroyed and with them their data when an an earth quake happens (and the ones nearest to the quake with the most interesting data are the most vulnerable ones; and their data cannot be examined later). To solve the problem there could be a backup system distributed over several nodes. Possible problems:

  • High amounts of data (each node produces 4 MB per hour)
  • backup traffic must be very low priority, because an earth quake can always happen (early warning detections)

Archiving system with MiniSeed export

The nodes have mainly two tasks:

  1. early warning detections of earth quakes
  2. archiving the data and export it in MiniSeed format

For the second part there already exists the SeisComP system, which needs to be ported to OpenWrt and a plugin needs to be created for the Sensorboard.

Links & Literature:

This task could also be split up into more sections.

Simulator coupling

In the SAFER project we use three different simulators: A home-made simulator for the sensor data analysis, ODEMx for the distributed event algorithms and ns-2 for the wireless part. How to couple these simulators (especially ns-2 and the home-made simulator) is still an open issue.

Analysis of Atheros & Linux WiFi Behaviour

The simulation of wireless networks is based on assumptions about the timing behaviour of each component. However, that very much depends on the hardware & software used. The goal of this project is to evaluate the timing behaviour of both the Atheros hardware, the device driver and the 802.11 stack used in the SAFER and EDIM projects. One result could be a ns-2 simulation of the components.

Distributed Software Deployment and Updates in a Wireless Mesh Network

Every system needs new or updated software from time to time, in development even regularly, and no one wants to update software by hand, especially in a wireless mesh network where you would have to update many systems. Thus an automated distributed software deployment and update scheme for a wireless mesh network like in the EDIM project is needed, which has to solve following questions:

  • How to distribute software efficiently?
  • How to ensure all nodes in the network received the software to update?
  • How to ensure that all nodes update at the same time (as they may become incompatible to each other with the update) ?
  • How to ensure version consistency?
  • How to recover from a bad update where something went wrong ?
  • (Special to our hardware: Is it possible to update the complete system, or only some software on top of the system?)
  • ...

See also SAR-PR-2006-01 Development of a Software Distribution Platform for the Berlin Roof Net (Diplomarbeit / Masters Thesis). Bernhard Wiedemann. 73 pages. or Roonk