1 / 29

A Cluster-based Approach for Data Handling in Self-organising Sensor Networks

A Cluster-based Approach for Data Handling in Self-organising Sensor Networks. UCL SECOAS team: Dr. Lionel Sacks, Dr. Matt Britton Toks Adebutu, Aghileh Marbini, Venus Shum, Ibiso Wokoma Presented by Venus Shum Advance Communications and Systems Engineering group University College London

kitty
Télécharger la présentation

A Cluster-based Approach for Data Handling in Self-organising Sensor Networks

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. A Cluster-based Approach for Data Handling in Self-organising Sensor Networks UCL SECOAS team: Dr. Lionel Sacks, Dr. Matt BrittonToks Adebutu, Aghileh Marbini, Venus Shum, Ibiso Wokoma Presented by Venus Shum Advance Communications and Systems Engineering group University College London Supervisor: Dr. Lionel Sacks

  2. Content • The SECOAS sensor network • Objectives and approaches of data handling • Spatial algorithms • Supporting platform and message exchange

  3. The SECOAS Sensor Network

  4. SECOAS project • SECOAS – Self-Organised Collegiate Sensor Network Project • Aim: To collect oceanographic data with good temporal and spatial resolution • Why SECOAS? • Traditionally done by 1 (or a few) expensive high-precision sensor nodes • Lack of spatial resolution • Data obtained upon recovery of sensor nodes • Equipment needs to be recovered at the end of the data gathering exercise • Burst data - May miss interesting features 1 2 3 4

  5. The sensor network approach • A distributed system/ network • Characteristics: • Large number • Low cost • Low processing power • Advantages • Provide temporal and spatial resolution • Data dispatched to the scientist in regular interval • Wireless ad hoc network • Stringent battery requirement • communication constraint 1 2 3 4

  6. SECOAS Specialties • Distributed Algorithms • A clustering approach for data handling • Biologically-inspired algorithms • A custom-made kind-of OS (kOS) tailor for implementation of Distributed algorithms 1 2 3 4

  7. Node architecture 1 2 3 4

  8. Network scenario 1 2 3 4

  9. Network scenario 1 2 3 4

  10. Objectives and approaches of data handling

  11. A simplified scenario • All nodes sample • Sampling • Temporal compression • Data route back to base station • Spatial compression when possible • Not optimal because • Data Redundancy • Power usage for sampling and comm. 1 2 3 4

  12. A clustering approach • A clustering approach for spatial data handling • the monitored area is partitioned into interesting groups • strategies are carried out based on the cluster formations. • Clustering Requirements specific to SECOAS • Scalable • Dynamic and adaptive • Simple • Distributed, not rely on underlying network architecture • Robust 1 2 3 4

  13. Resources analysis • Resources • Battery power + Processing power • Bandwidth • Memory • Data resolution is a goal • Abstract concept • set by user • Related to the environment 1 2 3 4

  14. A resource scenario • Data fusion save power, memory and bandwidth • Radio: processing = 20:1 in the first trial • Increase sampling nodes = increase resolution • Final results feedback to algorithms 1 2 3 4

  15. Parameter space • The parameters set  (physical phenomena of interest PPI) used for clustering • Need to find out what characterise the measurement – data analysis • Pressure, salinity, temperature, sediment, tilts • The Mean, does not mean a lot in most cases These all have the same mean! 1 2 3 4

  16. Spatial Algorithms

  17. Information Flow 1 2 3 4

  18. Auto-location algorithm • Iterative averaging • Position aware nodes (PA) and position determining nodes (PD) • Position propagates from PAs to PDs. PDs use averaging to estimate position iteratively. • Simple, distributed and self-organised 1 2 3 4

  19. Results - Auto-location 1 2 3 4

  20. Clustering Algorithm • An algorithm inspired by Quorum sensing carried out by bacteria cells to determine when there is minimum concentration of a particular substance to carry out processes such as bioluminescence. • Analogy • Concentration of substance => PPI • Bacteria cell => sensor nodes • Processed group => clusters • The range of the grouping is determined by LALI used by e.g. ant cemetery construction • LALI (local activation lateral inhibition) 1 2 3 4

  21. Results - clustering • Only local/ neighbour information is required for forming clusters. • Independent of topology • Dynamic and scalable 1 2 3 4

  22. Supporting platform and message exchange

  23. kOS – kind-of operating system • Full support of distributed algorithms • Individual algorithms responsible for scheduling their actions • Virtualisation of algorithms – • software can use kOS functions disregarding their physical location • Interfaces to other physical boards are provided • Easy exchange of parameters between algorithms • Adaptive scheduling to distribute resources according to environmental condition 1 2 3 4

  24. Interaction of algorithms within a node 1 2 3 4

  25. Parameter sharing among neighbours • Enable exchange of information between nodes • An interesting facts of UCL SECOAS team: • Consist of 4 (pretty) women and 1 guy => gossip! • 2 characteristics of gossiping • Selective/random targets • Don’t always pass information that is exactly the same! (Add salt and vinegar) 1 2 3 4

  26. Gossip protocol in SECOAS • Type 1: Passing the exact parameters to randomly selected nodes (multi-hop) • Type 2: Passing altered parameters to all neighbour nodes (also, one hop only) • Efficient protocol and avoid flooding • Low latency requirement and network has weak consistency 1 2 3 4

  27. Finally…

  28. Conclusion and Future work • SECOAS data handling uses cluster-based approach • Next step: • Find the suitable parameters (PPI) from data analysis • Investigate how they work with the clustering algorithm • Auto-location optimises using number of position aware nodes, signal strength, etc. • Investigate temporal compression and spatial fusion strategy

  29. Thanks for the attention! Now Q&A

More Related