Harsh Shukla - Academia.edu (original) (raw)

Papers by Harsh Shukla

arXiv (Cornell University), Sep 17, 2021

I have been influenced my many people in the time it took to complete the work in these pages. Fi... more I have been influenced my many people in the time it took to complete the work in these pages. First and foremost has been my advisor and mentor, Prof. Spandan Roy. I was fortunate enough to find someone who shares the same passion for self driving cars and how will they will shape our world in future. He helped me in pursuing the problems in this domain and guided me througout the time. Many of the ideas in this thesis were developed at his suggestion. I hope I have done them justice. I have had the pleasure of working in the Robotics Research Center with some great people. I met here some of the brightest minds seen. All these people inspired me and still inspire to do more and reach out for great things. They made the lab a fun and enjoyable place to talk about research, or anything at all. Special thanks to Mithun Nallana and Saransh Dave for their valuable inputs. Finally, I want to thank my parents and my brother for their patient support throughout all these years. I would like to thank people known me and have always been with me from my undergrad days. I would like to thank Anjali and Dheeraj, for always believing in me. v

arXiv (Cornell University), Oct 2, 2020

In many machine learning applications, one wants to learn the unknown objective and constraint fu... more In many machine learning applications, one wants to learn the unknown objective and constraint functions of an optimization problem from available data and then apply some technique to attain a local optimizer of the learned model. This work considers Gaussian Processes (GPs) as global surrogate models and utilizes them in conjunction with derivative-free trust-region methods. It is well known that derivative-free trustregion methods converge globally-provided the surrogate model is probabilistically fully linear. We prove that GPs are indeed probabilistically fully linear, thus resulting in fast (compared to linear or quadratic local surrogate models) and global convergence. We draw upon the optimization of a chemical reactor to demonstrate the efficiency of GP-based trust-region methods.

Computers and Electrical Engineering

The COVID-19 disease, initially known as SARS-CoV-2, was first reported in early December 2019 an... more The COVID-19 disease, initially known as SARS-CoV-2, was first reported in early December 2019 and has caused immense damage to humans globally. The most widely used clinical screening method for COVID-19 is Reverse Transcription Polymerase Chain Reaction (RT-PCR). RT-PCR uses respiratory samples for testing, because of which, this manual technique becomes complicated, laborious and time-consuming. Even though it has a low sensitivity, it carries a considerable risk for the testing medical staff. Hence, there is a need for an automated diagnosis system that can provide quick and efficient diagnosis results. This research proposed a multi-scale lightweight CNN (LMNet) architecture for COVID-19 detection. The proposed model is computationally less expensive than previously available models and requires less memory space. The performance of the proposed LMNet model ensemble with DenseNet169 and MobileNetV2 is higher than the other state-of-the-art models. The ensemble model can be integrated at the backend of the smart devices; hence it is useful for the Internet of Medical Things (IoMT) environment.

The tiger, a poster child for conservation, remains an endangered apex predator. Continued surviv... more The tiger, a poster child for conservation, remains an endangered apex predator. Continued survival and recovery will require a comprehensive understanding of their genetic diversity and the use of such information for population management. A high-quality tiger genome assembly will be an important tool for conservation genetics, especially for the Indian tiger, the most abundant subspecies in the wild. Here, we present high-quality near-chromosomal genome assemblies of a female and a male wild Indian tiger (Panthera tigris tigris). Our assemblies had a scaffold N50 of >140□Mb, with 19□scaffolds, corresponding to the 19 numbered chromosomes, containing 95% of the genome. Our assemblies also enabled detection of longer stretches of runs of homozygosity compared to previous assemblies which will improve estimates of genomic inbreeding. Comprehensive genome annotation identified 26,068 protein-coding genes, including several gene families involved in key morphological features such ...

Oxford Textbook of Suicidology and Suicide Prevention

Suicidal behaviours still remain one of the hardest challenges for clinicians and researchers wor... more Suicidal behaviours still remain one of the hardest challenges for clinicians and researchers worldwide. In 2006, a Section on Suicide and Suicide Prevention was established by the European Psychiatric Association General Assembly, whose name was later changed in 2007 into the EPA Section on Suicidology and Suicide Prevention (EPA-SSSP). The original impulse for the creation of the Section was to offer a logistic frame and reference to the scientific activity of European psychiatrists. Since then, the EPA-SSSP has had an ongoing growth process. Currently, the main Section’s activities include: congress and educational activities (in the context of the annual EPA congress as well as of other congresses of national, European, and worldwide importance); Scientific activities (developing specific guidelines for the treatment of suicidal patients, sharing knowledge about suicide and suicide prevention, and fostering debate among researchers); research activities (creation of networks, ai...

2020 4th International Conference on Automation, Control and Robots (ICACR)

A novel low-cost robotic solution is proposed to automate the library inventory management proces... more A novel low-cost robotic solution is proposed to automate the library inventory management process of book arrangement with little to no human intervention. A line following robot with multiple sensors and actuators has been designed to operate on a path connecting bookshelves and the drop location. A CAD model for the robot is designed and various subsystems have been described in detail. The riser subsystem has been simulated to evaluate the design parameters with a desirable factor of safety. The working principle of the robot and control algorithms involved in path planning and locomotion has then been discussed through the electronic architecture and software stack sections. A simple PID control scheme for lower-level actuation has been used to control the four traction motors. The feasibility of the proposed scheme is then validated with the development of a simple pilot prototype.

Annals of the Romanian Society for Cell Biology, May 23, 2021

2016 European Control Conference (ECC), 2016

In this paper, a novel optimality-tracking algorithm for solving Economic Nonlinear Model Predict... more In this paper, a novel optimality-tracking algorithm for solving Economic Nonlinear Model Predictive Control (ENMPC) problems in real-time is presented. Developing online schemes for ENMPC is challenging, since it is unclear how convexity of the Quadratic Programming (QP) problem, which is obtained by linearisation of the NMPC program around the current iterate, can be enforced efficiently. Therefore, we propose addressing the problem by means of an augmented Lagrangian formulation. Our tracking scheme consists of a fixed number of inexact Newton steps computed on an augmented Lagrangian subproblem followed by a dual update per time step. Under mild assumptions on the number of iterations and the penalty parameter, it can be proven that the sub-optimality error provided by the parametric algorithm remains bounded over time. This result extends the authors' previous works from a theoretical and a computational perspective. Efficacy of the approach is demonstrated on an ENMPC example consisting of a bioreactor.

Transposable elements (TEs) are selfish genomic parasites that increase their copy number at the ... more Transposable elements (TEs) are selfish genomic parasites that increase their copy number at the expense of host fitness. The “success,” or genome-wide abundance, of TEs differs widely between species. Deciphering the causes for this large variety in TE abundance has remained a central question in evolutionary genomics. We previously proposed that species-specific TE abundance could be driven by the inadvertent consequences of host-direct epigenetic silencing of TEs—the spreading of repressive epigenetic marks from silenced TEs into adjacent sequences. Here, we compared this TE-mediated “epigenetic effect” in six species in the Drosophila melanogaster subgroup to dissect step-by-step the role of such effect in determining genomic TE abundance.We found that TE-mediated spreading of repressive marks is prevalent and substantially varies across and even within species. While this TE-mediated effect alters the epigenetic states of adjacent genes, we surprisingly discovered that the tran...

International Journal of Adaptive Control and Signal Processing, 2021

Steer‐by‐wire (SBW) systems are considered as one of the most significant innovations among the t... more Steer‐by‐wire (SBW) systems are considered as one of the most significant innovations among the technologies developed for advanced driver‐assistance systems and autonomous vehicles. The main control challenge in a SBW system is to follow the steering commands in the face of parametric uncertainties and external disturbances; crucially, perturbations in inertial parameters and damping forces give rise to state‐dependent uncertainties, which cannot be bounded a priori by a constant. However, the state‐of‐the‐art control methods of SBW system rely on a priori bounded uncertainties, and thus, become inapplicable when state‐dependent dynamics become unknown. This work, to the best of the authors' knowledge for the first time, proposes an adaptive control framework that can tackle the state‐dependent uncertainties and external disturbances in a typical SBW system without any a priori knowledge of their structures and of their bounds. The stability of the closed‐loop system is studied analytically via uniformly ultimately bounded notion and the effectiveness of the proposed solution is verified via simulations against the state‐of‐the‐art solution.

Smith-Waterman is a well-known local sequence alignment algorithm that is used for finding region... more Smith-Waterman is a well-known local sequence alignment algorithm that is used for finding regions of maximum similarity between two biological sequences and is known to be a highly compute intensive task. As it is based on dynamic programming it guarantees optimal results. But Dynamic Programming has its own drawbacks such as heavy memory consumption and significant amount of computations. Many academicians and researchers have tried variety of methods to harness the large amount of computational capabilities provided by the GPU in order to make this algorithm run faster. This paper proposes a version of Parallel Scan Smith-Waterman algorithm to improve performance of its phase-2. Here, we have also compared and evaluated performance of proposed work with other approaches like anti-diagonal and blocked anti-diagonal for both constant gap model and affine gap model and have observed remarkable performance gain.

This paper describes the mechanical, electronic and software designs developed by Kharagpur RoboS... more This paper describes the mechanical, electronic and software designs developed by Kharagpur RoboSoccer Students’ Group (KRSSG) team to compete in RoboCup 2018. All designs are in agreement with the rules and regulations of Small Size League 2018. Software Architecture implemented over Robot Operating System(ROS), trajectory planning and velocity profiling, dribbler/kicker design and embedded circuits over the last year have been listed.

The link between economic growth and employment generation had been testified in the post reform ... more The link between economic growth and employment generation had been testified in the post reform period. The research paper attempted to scrutinise the variation in unemployment and economic growth in the recent past. the data had been collected from Centre for Monitoring Indian Economy, Reserve Bank of India, NSSO rounds and Periodic Labour Force Survey for the study. It was observed that LFPR gradually declined for females in a greater proportion then males. Further, LFPR was higher in rural areas then the urban areas. The lower female LFPR is cause of concern. The recent economic slowdown was an amalgamation of various labour market and money market variables. The down turn in economic growth started from 2018-19. The structural changes of demonetisation and GST affected the GDP growth in the short run while global economic slowdown affected the long run growth. The domestic savings were not able to support the investment which forced the use of foreign savings to boost the gr...

The research community has been making significant progress in hardware implementation, numerical... more The research community has been making significant progress in hardware implementation, numerical computing and algorithm development for optimization-based control. However, there are two key challenges that still have to be overcome for optimization-based control to be a viable option in the context of advanced industrial applications. First, the large existing gap between algorithm development and its deployment on platforms used by practitioners in industry. Second, from a more theoretical viewpoint, the lack of robustness of certain approaches, which are based on the unreasonable assumption that the model at hand perfectly represents the object under investigation. This thesis addresses the aforementioned challenges by establishing software toolboxes for automatic code generation, and proposing a data-driven methodology to enhance the performance of real-time optimization strategies during operation. The first part of this thesis focuses on the efficient implementation of Model Predictive Control (MPC) based on first-order operator splitting methods. Because of the cheap numerical operations associated with them, splitting methods are favorable candidates for applications with limited computing power. We first identify the computational bottlenecks and, subsequently, discuss their efficient deployment on processors, Field Programmable Gate Arrays (FPGA), and heterogeneous platforms. For rapid prototyping and deployment, two code generation toolboxes are developed: SPLIT and LAFF. These possess a high-level parsing interface for MATLAB and yield optimized C code that can be directly used in a variety of FPGA platforms. Features such as pipelining, memory partitioning, and parallelization are automatically incorporated, not requiring users to have in-depth knowledge about computer architecture and low-level programming. We then propose a framework to a priori solve the co-design problem arising in splitting method-based MPC to provide trade-offs between resources and latency. We provide analytical expressions that can avoid the daunting and time-consuming task of exploring the design space manually, thus reducing the final application development time.

We devise a new adaptive-robust control framework for tracking the control problem of a class of ... more We devise a new adaptive-robust control framework for tracking the control problem of a class of uncertain systems having state-dependent uncertainty and under the influence of time-varying input delay. In comparison to the existing adaptive-robust control (ARC) strategies, the proposed ARC framework removes the conservative assumption of a priori bounded uncertainty. In addition, the Razumikhin-theorem-based stability analysis allows the proposed scheme to deal with arbitrary variation in input delay. The effectiveness of the proposed ARC is verified via simulations and experimentations using a wheeled mobile robot demonstrating improved tracking accuracy compared to the state of the art.

Increasing habitat fragmentation leads to wild populations becoming small, isolated, and threaten... more Increasing habitat fragmentation leads to wild populations becoming small, isolated, and threatened by inbreeding depression. However, small populations may be able to purge recessive deleterious alleles as they become expressed in homozygotes, thus reducing inbreeding depression and increasing population viability. We used whole genomes sequencing from 57 tigers to estimate individual inbreeding and mutation loads in a small-isolated, and two large-connected populations in India. As expected, the small-isolated population had substantially higher average genomic inbreeding (FROH=0.57) than the large-connected (FROH=0.35 and FROH=0.46) populations. The small-isolated population had the lowest loss-of-function mutation load, likely due to purging of highly deleterious recessive mutations. The large populations had lower missense mutation loads than the small-isolated population, but were not identical, possibly due to different demographic histories. While the number of the loss-of-f...

2018 European Control Conference (ECC), 2018

In the context of static real-time optimization, the use of measurements allows dealing with unce... more In the context of static real-time optimization, the use of measurements allows dealing with uncertainty in the form of plant-model mismatch and disturbances. Modifier adaptation (MA) is a measurement-based scheme that uses first- order corrections to the model cost and constraint functions so as to achieve plant optimality upon convergence. However, first-order corrections rely crucially on the estimation of plant gradients, which typically requires costly plant experiments. The present paper proposes to implement real-time optimization via MA but use recursive Gaussian processes to represent the plant-model mismatch and estimate the plant gradients. This way, one can (i) attenuate the effect of measurement noise, and (ii) avoid plant-gradient estimation by means finite- difference schemes and, often, additional plant experiments. We use steady-state optimization data to build Gaussian-process regression functions. The efficiency of the proposed scheme is illustrated via a constrained variant of the Williams-Otto reactor problem.