Research Work

My research interests evolved over time as one field led me to other and motivated me to explore how do people conceive of developing the magnificent engineering solutions we see around at the first place! Along my journey, I met amazing set of peers, colleagues, mentors and lab mates, each with a different set of skills and ideas. Every project and person taught me something new. However, the most amazing part of it all has been how the boundaries and gaps between different theories, subjects and departments erode away and team work, determination and passion become the deciding factors for successful research.

Ongoing research

  • Near Optimal Network Design for Coflows [Jun’17-present]
    Mentor - Prof. Rachit Agarwal
    Dept. of Computer Science, Cornell University
    A network design has been proposed that provides near theoretically optimal average completion times for coflows – a networking abstraction for cluster computing frameworks. This is achieved by decoupling coflow scheduling from rate allocation to individual flows within a coflow. The proposed mechanism uses a novel weight-scaling technique to order coflows, and also shows that flows within a coflow can be scheduled in an arbitrary manner while maintaining the obtained ordering. Theoretical proofs and extensive simulations have been performed to ensure that the above mechanisms are sufficient to achieve near-optimal performance.
    Paper” accepted in ACM SIGCOMM 2018!

Undergrad research

I was part of the Mobile Communications Lab, IITK working under the guidance of Prof. A. K. Chaturvedi on various research problems on MIMO systems. Apart from Wireless Mobile Communication Systems, my work spanned across various areas including Machine Learning, Computer Networks, Wireless Sensor Networks, Signal Processing, Control Systems and Computer Vision, etc. Following is the list of research projects I've been associated with:

  • Low Complexity Detection using Likelihood Based Tree Search in Large MIMO Systems (UG Research Project - Fall’ 16) [Aug’16-present]
    Mentor - Prof. A. K. Chaturvedi
    Dept. of Electrical Engineering, IIT Kanpur
    A detection algorithm for Large MIMO systems has been derived using a Tree Search based algorithm coupled with a novel Error Likelihood metric for branching and emplyoing an Interior Point based QP solver. The algorithm provides near-optimal results and is more accurate and efficient than most of the available low complexity detection techniques. The error performance and complexity results have been verified numerically by numerous simulations in MATLAB for various antenna dimensions and QAM constellation sizes.
    Paper accepted in IEEE Wireless Communications Letters (Vol 6 Issue 4)!

  • A Critique of Network Distance Estimation Using Matrix Factorization (Term Paper for Digital Communication Networks Course)[Aug’16-Nov’16]
    Mentor - Dr. Ketan Rajawat
    Dept. of Electrical Engineering, IIT Kanpur
    A critical analysis was done for a paper proposing an algorithm for network distance prediction using matrix factorization. After studying the algorithm and its advantages, it was implemented in MATLAB for reproduction and verification of results. The arguments presented were checked and weak claims were pointed out. Simulations were also carried out for scenarios which were not considered in the paper thus confirming whether the algorithm was applicable in the experimentally unverified cases of networks.

  • Online Learning of Optimal Sensor Sampling policy for Wireless Sensor Networks (Summer Research Project) [May’16-Jul’16]
    Mentor - Prof. Bhaskar Krishnamachari
    Dept. of Electrical Engineering, University of Southern California, USA
    An online learning algorithm was proposed and simulated to learn the optimal sensing policy for wireless mobile devices which minimizes average user state estimation error under a given energy budget constraint. The system was modelled as a Markov Decision Process and the problem was reduced to solving an equivalent Linear Program. The simulations were carried out in Python using CVXPY library and showed feasible regret as compared to the genie. The mathematical proof of the applicability and regret bounds of the algorithm were then derived.

  • Performance Analysis of Dual Hop MIMO Mixed FSO/RF Systems (UG Research Project - Spring’16) [Dec’15-Apr’16]
    Mentor - Prof. A. K. Chaturvedi
    Dept. of Electrical Engineering, IIT Kanpur
    The performance analysis results were found for dual hop MIMO mixed free space optical/radio frequency (FSO/RF) systems. The expressions for outage probability and average bit error rate for various modulation schemes were derived in the form of generalized infinite power series using amplify-forward and channel-state-information (CSI) assisted relaying. MATLAB simulations were carried out to verify the accuracy of the results.

  • ECG Signal Analysis for Classification of Cardiovascular Diseases [Jul’15-Dec’15]
    A qualitative analysis of the ECG data using complex Gaussian wavelets to investigate the multi-scale, self-similar behavior and deviation via phase plots of the wavelet cross spectrum of ECG signals was performed in order to classify various cardiovascular diseases. ECG signals were further analyzed using S transform to overcome the limitations of continuous wavelet transform and make the results more consistent and reliable.
    Paper accepted in International Conference on Advances in Computing, Communications and Informatics – 2016!

  • Vision Based Control of a Quadcopter [May’15-Jul’15]
    Mentor-Dr. S. R. Sahoo
    Dept. of Electrical Engineering, IIT Kanpur
    The dynamics of a Quadcopter were studied and modelled. A PID Control System was developed for controlling the yaw-pitch-roll of the Quadcopter. The modelling, simulations and analysis were performed using MATLAB and Simulink. Dronekit API was used to communicate the Quadcopter autopilot with a companion computer over an RF link with MAVlink protocol. Vision algorithms were developed to identify a given size and color marker. Commands coded in Python were sent from the computer to the flight control board.

  • Vehicle Trajectory Prediction using a Catadioptric Omnidirectional Camera (SURGE-2015 Research Project) [May’15-Jul’15]
    Mentor-Prof. K. S. Venkatesh
    Dept. of Electrical Engineering, IIT Kanpur
    A practical method was developed to predict the future spatial temporal trajectories of vehicles at road intersections using a catadioptric omnidirectional camera equipped with an equiangular mirror. Tracking was done using the Camshift algorithm running alongside a Kalman filter to handle occlusions. Domain transformation from image space to real world was done using a geometric model. A computationally effective model for future trajectory prediction was developed and simulated with available video dataset.
    Paper accepted in International Conference on Advances in Computing, Communications and Informatics – 2016!