Hello, I'm Istiak Mahmud. I am currently pursuing a Ph.D. in Electrical Engineering at the University of North Dakota. I am also working as Graduate Research Assistant in Communications, Artificial Intelligence, Systems, and Security (CAISS) Lab at University of North Dakota.

Contact:

Phone No : +1 701 8852344

E-mail : istiak.mahmud@und.edu

GitHub LinkedIn Google Scholar

Education

University of North Dakota (Aug 2024 - Present)

PhD in Electrical Engineering

University of North Dakota (Aug 2024 - Fall 2026)

Master's of Science (MSc) in Electrical Engineering

Professional Experience

University of North Dakota (Sep 2024 - Present)

Graduate Research Assistant & Graduate Teaching Assistant

Walton Digi-Tech Ind. Ltd. (Mar 2021 - Jul 2024)

Research and Development Engineer (Senior Assistant Director)

Technical Skills

Programming Languages: Python, C++, C, Matlab, HTML/CSS, Arduino

Quantum: Qiskit, Cirq, QuTiP, IBM Quantum Lab, QAOA Simulation, Ansys Lumerical

PCB and 3D Design: Altium Designer, SolidWorks, Proteus, Sprint Layout, Saturn PCB

Firmware: STM32CubeIDE, Arduino, Keil

Developer Software Tools: Pycharm, Google Colab, Matlab, VS Code, Simulink, Bitbucket, Github

Developer Hardware Tools: Arduino, Raspberry Pi, ESP32 and 8266, STM32 Nucleo board, ATtiny

Editors: Latex, MS Office, Libre-Office, Photoshop CC

Others: Machine Learning, Deep Learning, OpenCV, Data Science, Image Processing, Article Writing, IoT, VLSI and Verilog

Research Interests

Quantum Computing, Quantum Communication, Quantum Machine Learning, Quantum Cryptography, Quantum Networks, Quantum Optics, Mobile Robotics, Multi Robot Systems, Image Processing, OpenCV, Embedded System, Artificial Intelligence, Machine Learning, Deep Learning, Schematic Design, IoT, Autonomous Systems, Reverse Engineering, Unmanned Aerial Vehicles

Recent Achievements

Best Poster Award

Our research, “JATNet: Intelligent Jamming Detection & Resilient Communication for Autonomous Vehicle Platoons,” secured 2nd Place 🥈 in the Best Poster Award at IEEE RTC 2026, held at Illinois Tech, Chicago.

Grateful to my advisor and co-author, Prof. Ahmed Abdelhadi, for his continuous guidance and support. Special thanks to the IEEE RTC organizers and committee for this recognition, and to the University of North Dakota Graduate School and AFRL for supporting my conference travel through travel grants.

Student Conference Travel Grant — U.S. Air Force Research Laboratory

Received the Student Conference Travel Grant from the U.S. Air Force Research Laboratory (AFRL) SDR University Challenge (2025–2026) to support the presentation of our research, “AI-Assisted Secure SDR-Based Vehicle Platooning with Real-Time Responses.”

Student Conference Travel Grant — UND Graduate School

Received the Conference Travel Grant from the School of Graduate Studies, University of North Dakota (2026) to support my participation and research presentation at IEEE RTC 2026 in Chicago, Illinois.

2nd Place – U.S. Air Force Research Laboratory SDR University Challenge 2024–25

Our team from the University of North Dakota secured 2nd Place 🥈 in the AFRL SDR University Challenge 2024–25 under the Signal Processing category!

Proud to be featured by MathWorks (MATLAB) and the Wright Brothers Institute (WBI) in official press releases for innovation in SDR and semantic communications. Special thanks to my amazing teammates and our advisor for their constant support. 💡

Best Paper Award – IEEE 5th Cyber Awareness and Research Symposium 2025

One of my papers received the 4th Best Paper Award at IEEE CARS, held from October 27–30, 2025, at the University of North Dakota, Grand Forks, ND, USA.

Best Paper Award – IEEE 4Th Intelligent Methods, Systems, and Applications (IMSA 2026) International Conference

One of my papers received the Best Paper Award at IEEE IMSA, held from July 11–12, 2026, at Modern Sciences and Arts University, Giza, Egypt.

Research Publications

Published

Artificial Intelligence in Quantum Communications: A Comprehensive Survey

Journal: IEEE Access

(As 1st author)

Doi:10.1109/ACCESS.2025.3585799

Abstract: The integration of Artificial Intelligence (AI) into the field of quantum communication has emerged in recent years as a promising avenue for enhancing secure data transmission, error correction, and scalability in quantum networks. This paper offers a comprehensive survey of AI applications in quantum communication, with a focus on machine learning (ML) models such as neural networks and reinforcement learning, which are adapted to manage complex quantum challenges. Despite the growing body of research, current studies often concentrate on specific AI techniques or quantum applications, lacking a unified framework that evaluates their practical effectiveness, scalability, and cost-efficiency. Addressing this gap, our survey systematically reviews and categorizes AI-driven methodologies across core aspects like system security, transmission speed, and reliability. Through a critical comparison and ranking of existing techniques, this study identifies best practices and key advancements while highlighting areas requiring further exploration. By providing an in-depth analysis of AI’s potential to transform quantum communication, this paper aims to serve as a foundational resource for researchers and practitioners seeking to develop resilient, adaptive, and economically viable quantum communication systems.

AI-Based Solution of Quantum Systems with Unknown Underlying Parameters

Conference: 2025 15th International Conference on Electrical Engineering (ICEENG)

Platform: IEEE Xplore (As 1st author)

Doi:10.1109/ICEENG64546.2025.11031325

Abstract: Quantum engineering is of crucial importance to the future of computing. At the heart of it is the Schrödinger equation. It is fundamental in quantum mechanics, describing the behavior of quantum systems such as the quantum harmonic oscillator. Solving such equations analytically or numerically can be challenging, particularly in the presence of noisy observations or incomplete knowledge of the system parameters. This research addresses these challenges by applying Physics-Informed Neural Networks (PINNs), an approach that integrates physical laws into machine learning frameworks. Using a fully connected feedforward neural network with dynamically weighted loss functions, we solve the Schrödinger equation for the quantum harmonic oscillator and reconstruct wavefunctions for quantum states n = 0, 1, 2 and 3. The proposed methodology balances physics loss, derived from the Schrödinger equation residual, and data loss from noisy observations. Results demonstrate excellent convergence of PINN-predicted solutions to exact wavefunctions, even in inaccurate measurements. This study contributes a robust framework for integrating physics and machine learning, with potential applications in complex quantum systems and experimental data analysis.

Physics-Informed Neural Networks for Quantum Wavefunctions

Conference: 2024 International Conference on Computer and Applications (ICCA)

Platform: IEEE Xplore (As 1st author)

Doi: 10.1109/ICCA62237.2024.10927810

Abstract: The wave functions and permitted energy levels of quantum systems, which characterize the probability distributions and particle behaviors, are able to be determined via the time-independent Schrodinger equation. Our research study introduces an experimental method of solving the time-independent Schrodinger equation by utilizing Physics-Informed Neural Networks (PINNs) with a customized loss function designed for output waveshape prediction. The technique makes use of PINNs' capacity to directly integrate physical rules into the learning process, assuring that the solutions follow the guiding principles of quantum mechanics. We forecast the associated wavefunctions for different energy levels (n states) and effectively solve the Schrodinger equation for distinct quantum states by using PINNs and a customized loss function. Our findings show that PINNs are a reliable and accurate way to represent the fine features of quantum wavefunctions, and they present a viable substitute for more conventional numerical techniques. This methodology not only improves computer performance but also offers more profound understanding of the behavior of systems in quantum mechanics, which may have implications for material science, nanotechnology, and quantum computing.

Neural Network Based Ray Tracing on a Digital Twin for Performance Approximation Using Parameter Analysis

Conference: 2024 International Conference on Computer and Applications (ICCA)

Platform: IEEE Xplore (As 1st author)

Doi: 10.1109/ICCA62237.2024.10927985

Abstract: Accurately estimating received power in wireless networks is essential for optimizing signal quality, coverage, capacity, and managing interference. This paper introduces a method that combines machine learning with ray tracing simulations in a digital twin environment to approximate performance in wireless systems. Using MATLAB and OpenStreetMap, data on various ray types: line of sight, reflected, refracted, and diffracted is gathered across different transmitter-receiver setups. Key features, like propagation delay, path loss, propagation distance, phase shift, number of interaction and receiver position, are fed into a multilayer neural network to predict received power. By integrating clustering techniques, this approach enhances prediction accuracy and reduces signal estimate errors, outperforming traditional models. Experimental results show the neural network's superior accuracy and efficiency, supporting real-time optimization in complex urban environments.

Comparative Analysis of ML Models for Power Prediction in Complex 3D Digital Twins Using Ray Tracing Environment

Conference: 2024 International Conference on Computer and Applications (ICCA)

Platform: IEEE Xplore (As 1st author)

Doi:10.1109/ICCA62237.2024.10927759

Abstract: In this study, we present the development of a comprehensive ray tracing dataset tailored for versatile 3D digital twin environments that closely mimic real-life scenarios. For precise modeling in complicated environments, this dataset includes crucial factors like phase shift, propagation distance, path loss, propagation delay, and the number of ray contacts or reflections. Using this dataset, we forecast received power and other relevant variables using a variety of machine learning (ML) techniques. To improve the data's suitability for machine learning models, we do feature extraction and exploratory data analysis to find patterns in the data. By comparing the performance and error metrics across different ML approaches, our analysis provides a robust framework for network developers seeking optimal ML strategies for intricate environments featuring real-life channels and objects. The results provide important information for choosing appropriate machine learning models for improved precision and effectiveness in ray tracing applications, ultimately promoting AI-driven solutions in network development and digital twin settings.

Advancing Wildlife Protection: Mask R-CNN for Rail Track Identification and Unwanted Object Detection

Journal: IEEE Access

(As 1st author)

Doi: 10.1109/ACCESS.2023.3313253

Abstract: Train incidents with animals and even humans have gotten more attention since the past. Every year, many animals are killed in train tracks, causing imbalance of ecosystem and significant delays in railway traffic. Similarly, sometimes the train accident is very common at the railgates, where the motor vehicles are crossing the train line. All that happens due to the lack of detecting the objects properly on the train line. And this detection is totally dependent on the driver or human, where the error happens sometimes to buzz the horn at the right time and the accident happens. That’s why there needs to be an automated solution to detect the objects on the train line and inform the driver about them. In this research, we proposed a system that’s able to detect objects only on the train line. Where some times there’s an object just beside the train line, which is in a safe distance, to eliminate this kind of wrong detection, we detected the train line first and then detected the objects on the detected train line. Here, we used the Mask R-CNN algorithm to detect the train line and objects on the train line. A railway traffic dataset with an input size of 512 x 512 pixels was used to test the proposed methodology, which came up with results with a mean average precision of 0.9375 and a frame rate of 30 frames per second. According to the findings of the experiments, the suggested approach may be applied to identify objects on railroads in the real world.

Cardiac Failure Forecasting Based-on Clinical Data Using A Lightweight Machine Learning Meta-model

Journal: Diagnostics - MDPI

(As 1st author)

Abstract:

Accurate prediction of heart failure can help prevent life-threatening situations. Several factors contribute to the risk of heart failure, including underlying heart diseases such as coronary artery disease or heart attack, diabetes, hypertension, obesity, certain medications, and lifestyle habits such as smoking and excessive alcohol intake. Machine learning approaches to predict and detect heart disease hold significant potential for clinical utility but face several challenges in their development and implementation. This research proposes a machine learning metamodel for predicting a patient’s heart failure based on clinical test data. The proposed metamodel was developed based on Random Forest Classifier, Gaussian Naive Bayes, Decision Tree models, and k-Nearest Neighbor as the final estimator. The metamodel is trained and tested utilizing a combined dataset comprising five well-known heart datasets (Statlog Heart, Cleveland, Hungarian, Switzerland, and Long Beach), all sharing 11 standard features. The study shows that the proposed metamodel can predict heart failure more accurately than other machine learning models, with an accuracy of 87%.

Doi: https://doi.org/10.3390/diagnostics13152540

An approach of cost-effective automatic irrigation and soil testing system

IEEE 2020 Emerging Technology in Computing, Communication and Electronics (ETCCE)

(As 1st author)

Abstract:

The population of Bangladesh reached beyond 16 Crores, and the rate of population is increasing every day. There will be a severe food issue after 25-30 years, so it is important to improve the agriculture. The farmers today suffer from a shortage of rainfall and water shortages. The key objective of this paper is to include an integrated irrigation network, and also to measure the quantity of the three major macronutrients, nitrogen (N), phosphorus (P), and potassium (K), in the soil, thereby saving the farmer's time, money, and strength. Comparing the solution to a color diagram determines the amount of N, P and K in the soil sample. It will describe the high, medium and low levels of N, P and K. Manual control is necessary for the conventional farmland techniques. Human interference can be minimized with automatic irrigation systems. This paper presents a smart system that will sense and analyze the NPK rate of soil and will detect the moisture rate of that land, then based on these values, automatic irrigation will happen. As a master of system ATtiny Microcontroller used here to make the device cost-effective and instead of pH meter here used TCS230 color sensor to detect the soil pH or NPK rate. To make the system more cost-effective, there used a Mobile app as the monitoring device instead of any external hardware display.

Doi: 10.1109/ETCCE51779.2020.9350896

Submitted

Sustainable Crop Planning: Integrating Embedded System and Machine Learning Meta-model for Improved Agricultural Yield

Journal: Frontiers in Plant Science

(As 2nd author)

Abstract:

Agriculture holds a prime position among the world's largest and most impactful sectors. Thus, optimizing agricultural output becomes pivotal for ensuring a nation's food security and the well-being of its population. A significant challenge within agriculture lies in accurately predicting the best crops for specific areas. Adding to this complexity is the rapid evolution of global climates. While progress has been made in agricultural research, it has predominantly concentrated on individual hardware or software aspects. Lamentably, there's a dearth of research concerning embedded systems—a holistic solution encompassing both hardware and software. This study addresses this gap by introducing a precise embedded solution for crop prediction, utilizing a composite machine learning meta-model that achieves an accuracy and precision rate of 97%. This meta-model amalgamates four distinct machine learning models. The hardware components are harnessed for measuring critical soil parameters like N, P, K levels, pH balance, humidity, and temperature. At its core, the system employs an ESP32 microcontroller along with specialized tools like a customized pH sensor, DHT11, and NPK Sensor. For practical application, an Android app featuring the meta-model has been developed. The ultimate goal is to empower the app to forecast the most fitting crops for a given land area.

Reviewer | Judge | Volunteer

Springer Nature:
1. Quantum Information Processing
2. Scientific Reports
3. The Journal of Supercomputing
IEEE Access
ND State Science & Engineering Fair
2025 ND State FIRST Lego League State Championship
IEEE 5th Cyber Awareness and Research Symposium 2025 ( CARS'25 )

BSc Thesis

Implementation of Internet of things: Real-time tracking system using Raspberry Pi

This is a real-time tracker approach by using Raspberry pi and NEO-6M GPS sensor. Mainly the project is objected for the Train or Railway tracking system and the food delivery workers. For firmware system there used Python and Matlab parallelly.

Projects

Electrical and Electronics

Touch Sensor with I2C

This is 12 touch pad containing touch sensor. This is supports I2C interface. This is special type touch pad PCB. Because this PCB also contains backlight LED for every touch pad. This sensor have 12 touch pads.

I also did it for without backlight LED and in square shape board outline.

Errasable Writing Tab Controller Board

Cholesteric liquid crystal display film reset after writing something on the film. The circuit is powered by a simple coin cell of 3V, reset by a tactile button, and has a lock slide SPST switch.

Smart Door Lock

The locker contains a 12 pins touch sensor designed by me, a fingerprint sensor, battery backup, a display for monitoring, a Hal sensor, etc. Here, used ESP32 MCU as the main controller unit

2.4GHz PCB Antenna

It can be used in all 2.4 GHz designs, especially where small space is required for the antenna. I make it open source design for developers on my GitHub account.

Rechargeable Fan

It's a rechargeable fan controller board using STM IC.

Aging

LED Bulb

LED 12W

LED 5W

Vehicle Tilting monitoring device for safe unloading of objects

This is ESP32 controlled device with CAN bus communications.

Featured Phone Printed Circuit Board

300MBPS Wi-Fi Router

This Router can also be used as 300Mbps. This router contains four RJ5 ports, a power push pull switch, a power port etc. This router contains three PCBs. One is main router, other one is display board and other one is LED indicator board. Left side picture is the multi-board project with these three PCBs.

NFC Business Card

I've designed an NFC antenna right here and used an IC with NFC capability, allowing to send both data and power. An LED with a 0603 SMD package was also utilized for wireless power indicator.

Artificial Intelligence, Machine Learning, Deep Learning, Robotics, Data Science, OpenCV and Internet of Things (IoT)

OpenCV Based Ripe Lemon Detection

I have developed a Ripe Lemon detection system using OpenCV, incorporating a custom model created by me. The custom dataset, constructed by myself, was utilized for training the model, employing Mask RCNN.

OpenCV - Eye Detection and Tracking

OpenCV, an open-source computer vision library, facilitates eye detection by employing image processing techniques to identify and locate eyes within a given image or video feed. Once detected, tracking algorithms can be implemented to monitor and follow the eyes' movements over successive frames, enabling applications such as gaze analysis and attentive user interfaces.

Face Detection using OpenCV

OpenCV, an open-source computer vision library, enables face detection through algorithms that analyze image patterns to identify and locate human faces within a given image or video stream. This capability is widely used in applications such as facial recognition, surveillance systems, and human-computer interaction.

Social Distance Monitoring by Computer Vision

Social distance monitoring by computer vision involves using algorithms to analyze images or video streams, detecting individuals, and measuring the distances between them in order to ensure adherence to recommended or specified social distancing guidelines. This application of computer vision technology is crucial for public health measures, providing a tool to monitor and enforce social distancing in various settings.

Line Follower Robot by PID Algorithm

The Line Follower Robot using PID Algorithm is an autonomous robot designed to follow a predefined path, typically a black line on a white surface. It uses infrared sensors to detect the line and applies a PID (Proportional-Integral-Derivative) control algorithm to adjust the speed of its motors, ensuring smooth and accurate turns. The PID controller continuously calculates the deviation from the line and corrects the robot’s path, resulting in stable and responsive motion even on curved or complex tracks.

Obstacle Avoidance Robot by ATmega MCU

The Obstacle Avoidance Robot using ATmega MCU is an autonomous mobile robot designed to detect and avoid obstacles in its path without human intervention. Powered by an ATmega microcontroller (such as ATmega328), the robot uses ultrasonic or infrared sensors to continuously scan its surroundings. When an obstacle is detected within a certain distance, the microcontroller processes the sensor data and commands the motors to stop, turn, or change direction to avoid collisions. This project demonstrates real-time decision-making and efficient use of embedded systems for safe navigation in dynamic environments.

Maze Solving Robot by PID Algorithm

The Maze Solving Robot using PID Algorithm is an intelligent autonomous robot designed to navigate and solve complex mazes efficiently. It uses infrared sensors to detect path boundaries and employs a PID control algorithm to maintain precise movement along the path. The PID controller minimizes errors while turning or following curves, enabling smooth and stable navigation. This robot intelligently makes directional decisions at junctions and continues exploring until it finds the maze exit, demonstrating the integration of control theory and real-time pathfinding.

IoT-Powered Smart Data Logging System for Retail Warehousing to Prevent Inventory Loss

Executed for a local e-store, this project addressed a crucial issue exacerbated by the Covid-19 pandemic. The challenge involved a lack of product records in the warehouse, leading to products being misplaced. To resolve this, I implemented a solution that logs data to a Google Sheet each time a load is added or released from a specific rack. The main controller is an ESP MCU, complemented by four load sensors to detect changes in load.

Smart home automation system

In this setup, an ESP32 is employed to manage home appliances, allowing control and status display from any locally connected device.

Smart Street Lighting using ESP32 and ThingSpeak

In this configuration, ThingSpeak is employed to oversee the light state. Additionally, an LDR facilitates automatic light control, and a PIR sensor detects moving objects. If the PIR detects movement at night, a mail is sent to the pre-saved address, even if the light remains off.

IoT based smart irrigation and soil testing system

In this system, the MCU autonomously regulates the water pump according to sensor data and provides a daily email to the pre-saved address, conveying information about the soil condition.

Brain Tumor Classification

A dataset comprising brain tumor information was utilized, encompassing both first and second-order features extracted from the images. This dataset includes five first-order features (Mean, Variance, Standard Deviation, Skewness, Kurtosis) and eight texture features, with the target level indicated in the "Class" column. The first-order features consist of Mean, Variance, Standard Deviation, Skewness, and Kurtosis, while the second-order features encompass Contrast, Energy, ASM (Angular Second Moment), Entropy, Homogeneity, Dissimilarity, Correlation, and Coarseness. The "mn" column represents image names, and the "Class" column indicates whether the image contains a tumor (1 = Tumor, 0 = Non-Tumor).

Heart Fail Forecasting by Machine Learning Meta-model

This research project has been published in the MDPI Diagnostics journal, and you can find comprehensive details in the publication section, including the abstract.

Suitable Crop Recommendation by Custom Machine Learning Model

The research project has been presented to the MDPI Agronomy journal, and you can access thorough information in the publication section, which includes the abstract.

Active Train Lane Detection by SSD

This represents another one of my published research works in IEEE Access. However, this particular project constitutes a segment of that published research, focusing on the utilization of the Mask R-CNN algorithm for train line detection.

3D Modeling by SolidWorks

SMT / Solderable Standoff Nuts - M3

Electrolytic Capacitor

Coil Inductor

CONN HEADER VERT 8POS 1.5MM_15SH-A-08-TS

Future Projects (Seeking for supervisor and Funding)

Conceptual Quantum Computer

I am developing a conceptual quantum computer, a low cost educational platform designed to demonstrate quantum computing principles using laser optics and simple electronics. Although it is not a real quantum computer, it will visually and interactively illustrate key concepts like superposition, interference, and measurement through polarization states, optical gates, and detectors. This system aims to serve as an affordable and practical learning tool for students and researchers in educational institutes to explore the fundamentals of quantum computing.

Re-settable specific customer following smart super-shop trolley.

Introducing an intelligent supermarket trolley robot capable of initiating training for a new customer with a button press, following with another button, and resetting for the next customer with a reset button. The implementation involves OpenCV and machine learning for training and testing, while a PID algorithm ensures the smooth operation of the robot.

WalkableTree

Smart and automated Tree-pot with IOT and Sensors Integration.

Page updated

Google Sites

Report abuse