Empirical prediction of packet transmission efficiency in bio-inspired Wireless Sensor Networks

Biological networks (specifically, genetic regulatory networks) exhibit an optimized sparse topology and are known to be robust to various external perturbations. We have earlier utilized such networks, particularly, the gene regulatory network of E. coli, for constructing smart communication structures in bio-inspired Wireless Sensor Networks (WSNs) having high packet transmission efficiency. In this paper, we present machine learning approaches to relate the graph topology based characteristics of such bio-inspired WSNs to their network-level robustness in terms of average packet transmission efficiency. In particular, we generate a support vector regression model using the graph metric features as input data. The model predicts the percentage of packets received by the highest degree sink node and a theoretical estimate for the overall network robustness.