Research Experience
Aug 2021 - Present
Penn State, PA, USA
ABOUT
THE RESEARCH
The proposed next-generation PAT system rests on a collection of innovations on the fronts of transcranial ultrasound, image reconstruction, optical engineering, and stretchable devices. The proposed efforts will push forward the frontier of functional human brain imaging. It addresses an unmet need for a relatively inexpensive imaging modality that can perform noninvasive functional imaging of the human brain with excellent spatial resolution without compromising the temporal resolution.
July 2022 - Present
Penn State, PA, USA
ABOUT
THE RESEARCH
This review discusses the recent developments and advantages of triggering strategies for controlling the degradation of on-demand transient electronics. We also summarize bioresorbable sensors for medical diagnoses, including representative applications in electrophysiology and neurochemical sensing. Along with the profound advancement in medical diagnosis, the commencement of therapeutic systems such as electrical stimulation and drug delivery for the biomedical or medical implant community has also been discussed. However, implementing a transient electronic system in real healthcare infrastructure is still in its infancy.
Mar 2022 - Present
Penn State, PA, USA
ABOUT
THE RESEARCH
This work presents a facile and low-cost fabrication method to integrate an ultrathin ionic layer with gradient microstructures with programmable profiles and heights created by a simple CO 2 laser. The coupled electrical and mechanical simulations provide a route to optimize the design of iontronic pressure sensors based on the electric double layer to address the existing challenges for significantly improved pressure sensing performance. The resulting optimized sensor exhibits a high sensitivity of 33 kPa − 1 over an ultra-board linear sensing range of 1700 kPa, an ultralow detection limit of 0.36 Pa, and a pressure resolution of 0.00725% under ultrahigh pressure of 2000 kPa.
Dec 2019 - Nov 2020
Jadavpur University, India
ABOUT
THE RESEARCH
The paper uses statistical and differential geometric motivation to acquire information about the learning capability of a neural network before its training. The analysis proves that there always exists a trade-off between parameter optimation rate and generalization capability of a neural network. During the training period, the system can equilibrate during some epochs, or the system can also tend towards non-equilibrium with an absence of diffusion. The Complexity-Action conjecture is shown by correlating learning trajectory with the trajectory of a particle in the gravitational field.
Aug 2018 - Aug 2021
Jadavpur University, India
ABOUT
THE RESEARCH
Resistive Random Access Memory (RRAM) was fabricated by dispersing amorphous Carbon Nanotubes (a-CNT) in a PMMA matrix. The paper analyzes the I-V characterization curve with the variation of the concentration of a-CNT. As the concentration increases, there is an increase in threshold voltage but with a decrease in the on-off ratio. Changing the variation of concentration spatially, an ultra-low threshold with high on-off RRAM has been fabricated. The system could be statistically mapped on to a spintronic system. Using the correlation, a simulation has been proposed using localization-favored NEGF formulation to fit the experimental data.
The paper presents a generalized framework of the surface topography by coarse graining the total asperities space in two coordinate system. The paper uses a generalized form of Majumdar and Tien (M-T) model which could also capture the hysteresis behaviour during loading and unloading using a variant model of the truncated Gaussian model. We extended the theoretical work to predict the nature of truncated area of topmost layer from an experimentally observed hysteresis favoured conductance-force plot using a neural network.
Aug 2020 - Nov 2020
A geometric perspective of Quantum Choas & Complexity in Quantum Neural Network
Germany
ABOUT
THE RESEARCH
The paper considers a hybrid quantum-classical framework of parameterized unitary function with stochastic gradient based learning algorithms. Using differential geometric and statistical approach, the paper makes an attempt to map the learning trajectory or unitary evolution of Quantum Neural Network (QNN) to the trajectory of particles in the parameter space using a Riemannian manifold called diffusion metric, proposed by Foressi. The paper establishes the correlation between the evolution of quantum chaos and complexity with the neural architecture of QNN.
April 2020 - May 2020
India
ABOUT
THE RESEARCH
The proposed method named Sewage Pooling Algorithm tests wastewater samples from sewage systems to pinpoint the regions which are affected by maximum chances of the virus spread. The algorithm also uses a priority-based backtracking procedure to perform testing in sewage links depending on the probability of infection in the sub-areas. For places with very rare CoVID cases, we present a gradient-based search method to prune those areas. The proposed method has less human intervention and increases the effective tests/million people over current in-place methods.
Jan 2020 - Mar 2020
Estimation of Pressure from Velocity Fields around a Two Dimensional NACA 4412 Airfoil using Artificial Neural Networks
Neptune Lab
Jadavpur University, India
ABOUT
THE RESEARCH
In the present paper, the conventional approach of solving pressure poisson equation is replaced by an artificial neural network (ANN). The flow over a NACA 4412 airfoil is considered for the present investigation. The required database for ANN training is generated using commercial CFD solver by solving Navier-Stokes equations. Three different neural networks have been developed based on the size of the training data set. The neural network gets trained in Reynolds number ranging from Re 50 − 500 with an interval of 50 and gets tested with intermediate Reynolds numbers.
Aug 2019 - Dec 2019
A generalized framework of Thermal Contact Conductance: A coarse-grain approach
Jadavpur University, India
ABOUT
THE RESEARCH
The paper presents a generalized framework of the surface topography by coarse graining the total asperities space in two coordinate system. The paper uses a generalized form of Majumdar and Tien (M-T) model which could also capture the hysteresis behaviour during loading and unloading using a variant model of the truncated Guassian model. We extended the theoretical work to predict the nature of truncated area of topmost layer from an experimentally observed hysteresis favoured conductance-force plot using a neural network.
June 2019 - Nov 2020
Fluid Inspired Path-planning for BCI-controlled Wheelchair
Jadavpur University, India
ABOUT
THE RESEARCH
The present work proposes a novel BCI framework of brain based shared-control strategy of wheelchair navigation by employing path planning algorithm using artificial potential theory inspired by fluid dynamics. The path planning algorithm considers the human confusion in target selection and provides an energy optimal path for the wheel chair in presence of dynamic obstacle. A hybrid BCI approach is taken here, which utilises SSVEP, MI and P300 brain response for the navigation of wheelchair, where the first modality is used for destination selection and rest of the modalities are used for local navigation.
June 2019 - Nov 2020
Jadavpur University, India
ABOUT
THE RESEARCH
The paper proposes a modification in the classical Q-learning by introducing a human agent in the learning loop. Such modification leads to remarkable improvement in the convergence speed and quality of solution. The near-energy optimal trajectory drawn by the human agent is discretized in small steps. This discretization plays a vital role in determining the instructor dependence in learning process. A linear variational inequality based primal dual neural network is employed here to find out the least energy joint combination from several combination generated by inverse kinematics. The trajectory having minimum torque requirement to perform the joint-orientation receives maximum reward in the Q-learning.
Education
2021 - Present
PENN STATE UNIVERSITY
DEPARTMENT OF ENGINEERING SCIENCE
STATE COLLEGE, PA, USA
CGPA: 4.0/4.0
2016 - 2020
JADAVPUR UNIVERSITY
DEPARTMENT OF MECHANICAL ENGINEERING
Kolkata, India
CGPA: 8.13/10.0
Research Topics
Dynamical Systems
Underactuated Robotics
Quantum Transport
Statistical Learning Theory
Reinforcement Learning
Deep Learning
Simulation + Programming
Python
MATLAB + SIMULINK
ROS
COMSOL