Design of large-scale IoT networks

Principal Investigator

Prof Rajesh Sundaresan (Professor, Robert Bosch Centre for Cyber-Physical Systems, Department of Electrical Communication Engineering)


A well-designed communication network that provides reliable quality of service (QoS) guarantees is a critical platform on which a variety of novel cyber-physical system (CPS) applications can be built. When designed well, the CPS entities can seamlessly talk to each other, and the network acts merely as a transparent enabler. When designed poorly, bottlenecks, congestions, delays, etc. that develop can choke not only CPS applications and but also the development of new CPS applications.

We are at a point in time where large scale IoT deployments, those that can improve delivery of societal services, are just beginning – smart meters at homes, safety monitoring of equipments, mobile health, improved distribution and leakage detection in water networks, among many others. Many innovative services are also being developed around such applications. It will not be long before IoT data starts flooding the network. It is therefore crucial that the right design decision choices are made at this stage, when the IoT networks are being planned, because once deployed, there is always reluctance to modify them.

IoT networks provide a significant challenge because end-nodes have limited power resources, for example, they may operate on batteries that are either never replaced or are charged via harvested energies bringing stochasticity into the picture. Such devices may also have to communicate over large distances in some (perhaps rural) regions, but may carry only sporadic (but usually light) traffic, and this sparsity must be intelligently exploited. Additionally, good deployments require tedious measurements by skilled engineers. The goal of this project is to identify IoT network design methodologies that can deal with the resource challenges highlighted above.

We have three specific objectives in mind.

  • Identify frequency agnostic, data-driven, automated network deployment strategies. This will help save costly network engineer’s time.
  • Design more robust physical layer technologies via coding that exploits polarisation diversity. This will allow for random deployments by subsequently enabling the network to self-heal and improve diversity order in case of random realisations that involve deep shadows.
  • Provide a model for deployment outcomes and for traffic, and identify system-optimal operating points. This will improve the QoS and overall throughput.

It is anticipated that this project will feed into the Smart Cities project’s goals.