Diagnosis of Sensor Networks

Jaikaeo, Srisathapornphat, Shen

sensor networks fault management diagnosis

@inproceedings{jaikaeo:icc-2001,
  title={Diagnosis of Sensor Networks},
  author={Jaikaeo, C. and Srisathapornphat, C. and Shen, C.C.},
  booktitle={{IEEE} International Conference on Communications},
  year={2001}
}

High nodes to managers problem may generate much management traffic

• Too many nodes to focus on individuals
• Centralized diagnosis node may become a bottleneck

Sampling
Self-orchestrated
Diffused computation

Hierarchical clustering
Attribute-based naming
Querying and tasking

Samplecast: Variation on flooding queries and all replying back to server in which nodes only reply with some given probability

• Adaptive Probabilistic Response: Refine even further, such that probabilities are different in different clusters
  • Provides for many responses from sparse clusters, less responses from dense clusters, in order to meet coverage goals
    • Not getting enough responses can significantly impair accuracy, a problem on sparse networks

Attempts to further reduce collissions by having nodes pick random response delay based on hop count to root, but presented scheme requires knowing how many nodes are at that depth...

Finding faulty sensors by finding those well out of whack with local readings from other nodes

• Cluster heads poll nodes, compare values to averages, report off kilter nodes to manager

Clusters are pre-configured