Delay Analysis of Data Collection in Sensory Networks - 03/17/2006

Masoud Sharif

In this talk, we study the delay in collecting (or distributing) data from sensors to the processing center via simple discrete mathematical models. Specifically given the number of packets in each node, we obtain the minimum delay to collect data achievable by using any centralized algorithm for various topologies such as line, multi-line and tree.

In the next part of the talk, we assume that the number of packets at each node is random and we obtain the average minimum delay in collecting data for a line or multi-line network. For large number of relays, we show that the same asymptotic behavior for the average minimum delay can be achieved by a simple decentralized algorithm. We also examine various parameters such as size of the packet, transmission range, and channel erasure probability on the expected delay.

This is based on a joint work with Cedric Florens and R. J. McEliece at Caltech.