Learning-based optimization strategies for sustainable IoT communications

About this project

Project description

In this project, we intend to develop the innovative data-driven techniques for application context aware smart computing and communication strategies at the edge nodes for 5th generation Internet of Things (5G IoT) and beyond communications. The use cases are important from both indoor as well as outdoor perspective, though the operating conditions, available resources, and objectives could be significantly different. Usually the smart devices/robots/sensors used in such use cases are of heterogeneous types and in large scale. Mobile communication and energy agents (e.g., unmanned aerial vehicles (UAVs)) are also deployed to undertake certain important tasks such as sensing, controlling, recharging, managing, remote diagnosing, etc.

To perform the desired tasks by the appropriate smart device and/or mobile agents and at the right time requires coordination among the devices/ mobile agents and the central control system. For this, data collected by sensors and mobile agents are to be communicated either to a central system or to set of distributed agents or nodes depending upon the use case and the architecture. Primary objective of this project is to study context-aware, sustainable communication and networking issues in deployment of smart sensors and mobile agents and to provide optimal solutions in terms of algorithms, protocols, and deployable proof-of-concepts on application aware network platform.


Some of the-problem areas in the project are:

  • Application context specific communication network protocol design for 5G IoT use cases. The contexts are expected to be stochastically dynamic. The specific cases of interest are: (i) smart city and industrial pollution map creation and localization and (ii) smart-grid/smart-metering applications.
  • In all use cases, while performing the desired task, the energy, bandwidth, and storage requirements are critical to resource efficiency. To this end, an effective strategy would address the problem by learning from the individual time series of sensed data as well as from inter-sensor coordination, which could be at the node-level as well as at the network-level.
  • Notwithstanding the efforts on energy efficiency, for uninterrupted network operation, a node with finite battery capacity needs to be recharged online. To this end, many industrial/ agricultural application scenarios would require dedicated wireless charging resource, where radio frequency (RF) energy transfer would be of interest. This is a crucial point from 5G prospective as alternate energy usage is a must item and through this it can be achieved to some extent. Mobile robots with terrestrial or aerial mobility will help while avoiding communication traffic flow related funnelling problem.
  • Specific interest would be drone/UAV-assisted energy replenishment and data collection, where the basic channel models may need to be re-investigated. Channel models are crucial for network architecture and deployment.

In the project scope primary focus would be given to learning and processing intelligence at the edge. Beyond data-specific features, effects of communication network constraints, e.g., congestion-induced delay and jitter, and wireless channel induced stochastic losses would be of interest in this project. Channel induced stochastic losses will be accounted as important factors from communication protocol and algorithm design and validation viewpoints. Context-specific network communication platforms will be developed, which can be adapted to satisfy various QoS requirements in 5G IoT use cases, namely remote connectivity, smart grid, and smart city. The two application domain specific broad classification of air-to-ground (AtG) communication studies would be: (i) urban/suburban, smart city type of sensing and automation applications, (ii) rural agricultural deployments for mechanized farming. For sustainable sensor node operation via on-demand wireless energy supply, RF energy transfer and our recently-proposed multihop RF energy routing technology will be refined by incorporating multi-antenna and beamforming technologies, which have not been explored before.


  • Multiple new concepts (e.g., architecture, protocols, algorithms, proof-of-concepts, and embedded system implementations) will come out of the project, which can also be used as standalone components.
  • The various context-specific IoT data, e.g., localized pollution samples and smart meter data from the selected appliances, will be made available on public repository for further analysis and research.
  • The research outcomes will be presented as papers in reputed international conferences, journals, and magazines.
  • Intellectual property (IP) filings on some of the novel learning-based algorithms are expected.

Information for applicants

Essential capabilities

Strong mathematical background, strong background in Signals and Systems and in Probability and Stochastic Processes.

Desireable capabilities

Interest/capability of working on hardware systems (implementation/experimentation using SDR/USRP kits and embedded hardware systems) will be advantageous.

Expected qualifications (Course/Degrees etc.)

The students are expected to have degrees in Electrical Communication/Computer Engineering or allied areas such as Applied Mathematics.

Candidate Discipline

Electrical Communication Engineering, Computer Engineering/Technology, Applied Mathematics, Signal Processing.

Project supervisors

Principal supervisors

UQ Supervisor

Associate professor Marius Portmann

School of Information Technology and Electrical Engineering
IITD Supervisor

Professor Swades De

Department of Electrical Engineering