Research scholars

Students are the best assets of a teacher for they are the ones who make him/her so. 

Here are mine.

My policy is to take scholars interested in research, not those interested in getting a doctoral degree!

Current Doctoral Candidates:

Past:

Currently: Machine Learning Scientist, Eli Lillie, Bengaluru

Thesis : Immuno-Inspired Embodied Lifelong Learning in Robotics 

The work describes the formulation of immuno-inspired mechanisms to continuously evolve, cache, manage and evict several Artificial Neural Network (ANN) based robot controllers, within disparate Halls-of-Fame, thereby facilitating Embodied Life-long learning in robots. The work also introduces a novel concept termed Mutational Puissance to enhance learning in ANN based controllers that use neuro-evolution. Further, unlike the conventional layer-wise transfers conducted in ANN-based Transfer Learning, a new immunology inspired Neuronal-level Transfer Learning technique has also been described. The technique aids in identifying neurons that play a more significant role during the learning phase. These, so-called, hot neurons, when transferred to target ANNs hasten learning convergence, especially when the source and target dataset domains are dissimilar. Transfer of such neurons has also proved to be effective while learning in scenarios involving robots 


Currently: Applied Scientist II, Microsoft, Hyderabad

Earlier: Post Doctoral Research Associate, Offshore Robotics for Certification of Assets (ORCA), University of Edinburgh

Thesis: On Decentralizing Intelligence in Cyber-Physical Systems

The thesis presents the making of a decentralized CPSs (dCPS) using an agent based model.A Location-Aware and Tracking Service (LATS) real-world dCPS based application is used to portray the viability of the proposed approach. A novel mechanism for ordering task executions to mitigate the issue of Mutual Exclusion of Shared Resources in a dCPS of multiple robots operating in a warehouse has been proposed and devised. The mechanism provides unique features such as addition and deletion of both tasks and robots, on-the-fly, without bringing the system down. The thesis also describes how a hybridized mechanism inspired by immunological models can select and find the best solutions in a decentralized manner. Finally, a mechanism to evolve solutions in the form of neural network based sub-controllers for sub-tasks to be executed by robots using a cross-reactivity threshold is presented.  

Fellowships Recieved: TCS Fellow

Travel Grants: Science & Engg. Research Board, Govt. of India, TCS Travel grant, SIAM Student Travel Award, Microsoft Research Travel Award


Current Position: 

Assistant Professor (Contractual), Dept. of Computer Science & Engg., National Institute of Technology New Delhi, India

Thesis: Non-intrusive Human Sensing – Techniques and Applications

The work reported in this thesis attempts to combine the information obtained from multiple heterogeneous sensors to improve the accuracy of non-intrusive human sensing. Investigations into dynamic environments have also been carried out. An immuno-inspired mechanism has been proposed, next In order to reduce redundancy in the set of features extracted thereby reducing computational costs. This mechanism was implemented using a Pyro Infrared (PIR) sensor and an Analog Ultrasonic Sensor (AUS). The thesis also describes an approach for early detection of cognitive impairment using spatio-temporal data obtained using human sensing. Since capturing such data for every scenario is not feasible, the thesis describes a manner in which transfer learning can be used to circumvent this issue.


Current Position:

Assistant Professor, Dept. of Computer Sc. &Engg., Indian Institute of Technology Ropar, Punjab, India

Thesis: On Mobile Agents for Learning & Coordination in a Networked Robotics Milieu

The work aims at embedding and emulating bio-inspired intelligence including Idiotypic Networks, learning and sharing algorithms, synchronization of asynchronous tasks and handling multiple and concurrent task allocations - all on a real network of nodes and robots comprising a Cyber-Physical system. Emulation is currently being carried out using Typhon, a mobile agent platform developed indigenously at the Robotics Lab. of the CSE Dept. at IIT Guwahati.

Achievements: One best paper award

Fellowships: TCS Fellowship

Travel Grants: Department of Science & Technology (DST), Govt. of India, TCS Travel grant


Samir Kr. Borgohain (2016)

Current Position:

Associate Professor, Dept. of Computer Sc. & Engg., National Institute of Technology Silchar, Assam, India

Thesis: An Immuno-inspired Technique for Language Generation & a Pictorially Grounded Language for Machine Aided Translation

The work describes the use of an Immune (Idiotypic) system for grammarless generation of correct sentences. An Idiotypic-like network, that functions much like the one Nobel Laureate N.K.Jerne described about Chomsky’s generative grammar, was realized that can generate new sentences, given a set/corpus of sentences. The network learns in a manner kin to how a small child learns a new language. A Pictorially Grounded Language that aids in removing semantic ambiguities to convey the right meaning in terms of pictures and animated images was also developed alongside to ensure grounding of symbols in the language.

 

W. Wilfred Godfrey (2012)

Current Position:

Associate Professor, Dept. of Computer Sc. &Engg., 

Indian Institute of Information Technology & Management, Gwalior, Madhya Pradesh, India

Thesis:  Mobile Agent based Bio-inspired Mechanisms for Servicing Networked Robots

The work portrays bio-inspired mechanisms that can guide a set of mobile agents towards those (networked) robots that require services carried by them as payload. The mechanisms make use of pheromone diffusion by robots and the conscientious and cloning behaviours on part of the mobile agents populating the network to culminate in a faster service. Both these entities(robots and mobile agents) together take part in a bi-directional search-and-serve exercise for resources over the network. An efficient pheromone mobile agent migration mechanism coupled with a controlled cloning strategy has also been described. The cloning control relies on stigmergy, thus saving bandwidth. The mobile agents inherently usea novel concept termed a cloning resource to exhibit stigmergy based cloning. Simulators for these mechanisms were coded. These mechanisms were also implemented on a set of real robots. The implementation revealed that such mechanisms can also be used to realize an Internet of Things (IoT). A proof of concept using real robots and sensors was also demonstrated to substantiate this claim.

02 patents