Md Al Maruf

Md Al Maruf

Assistant Professor of Computer Science and Mathematics at Algoma University.

Posts by Collection

experience

portfolio

ESOps: AI-Driven Embedded DevOps

A next-generation embedded DevOps research direction connecting feature discovery, testing, hardware-aware delivery, and assurance for long-lived embedded software.

Illustration of a software development and DevOps workflow
AI-assisted embedded DevOps workflow

Intelligent Grid and BESS Integration

An active research project exploring embedded optimization, real-time analytics, and digital-twin methods for resilient energy infrastructure and battery energy storage systems.

Illustration of an intelligent embedded controller platform
Embedded intelligence and control concept

Aircraft Landing Behaviour Detection Application

The Aircraft Landing Behaviour Detection Application stands as a pivotal innovation in air traffic control technology, setting new standards for safety, efficiency, and data-driven decision-making in airport operations. This project brings forth a paradigm shift in managing aircraft landings and runway statuses by providing real-time, actionable insights. Leveraging advanced algorithms and state-of-the-art sensor technology, the application ensures a seamless interface for air traffic controllers, enhancing their ability to monitor, predict, and act upon dynamic flight patterns and runway conditions.

Aircraft trajectory visualization
Aircraft trajectory visualization
Predicted aircraft flight path
Predicted flight path

Key Features:

  • Secure User Authentication: Ensures access is granted only to authorized personnel, maintaining the integrity and security of sensitive operational data.
  • Interactive Main Application Interface: Offers a user-friendly platform for monitoring real-time flight data and airport operations.
  • Detailed Runway Overview: Provides status updates for each runway, allowing for efficient traffic management.
  • Advanced Landing Detection: Utilizes cutting-edge technology for accurate landing insights.
  • Comprehensive Airline Trajectory Visualization: Enables precise tracking of flight paths for optimal route analysis.
  • Predictive Flight Path Analysis: Employs predictive algorithms for proactive air traffic management.

A Framework for Over-the-Air (OTA) Software Update

Fog computing emerges as a vital solution for time-sensitive vehicular over-the-air (OTA) updates, offering enhanced network durability and lower communication delays compared to the cloud. Our algorithm optimizes fog node resources, reducing OTA update times and improving network efficiency. The application of machine learning for communication delay prediction and the strategic enabling and disabling of fog nodes based on traffic load underscore the system’s innovative approach to OTA updates.

System integration middleware architecture
OTA update system-integration architecture

Key Features:

  • Optimized Fog Node Utilization: Efficient allocation of fog resources based on traffic patterns, reducing over-provisioning and associated delays.
  • Machine Learning-Driven Delay Prediction: Accurate prediction of communication delays between fog nodes and vehicles, ensuring timely updates.
  • Enhanced Resource Allocation: Strategic enabling and disabling of fog nodes to match traffic load, maximizing network responsiveness.
  • Comprehensive Delay Analysis: Includes handover and propagation delays, transmission rates, and vehicular mobility to predict OTA update time.
  • Real-World Data Validation: Utilization of European WiFi hotspot and 5G datasets to confirm the effectiveness of the proposed approach.
  • Scalability Assessment: Examination of throughput performance with varying vehicle numbers and OTA update sizes, ensuring robust system scalability.
  • Testbed Corroboration: Verification of simulation performance with a real-world testbed, employing QEMU virtualization and the Uptane framework.

Feature and Requirement Extraction Tool for Improving Embedded Software Reuse

The Feature and Requirement Extraction Tool is an advanced solution for analyzing embedded software, identifying significant functions, and extracting functional and non-functional requirements. Designed to work with .c files, this tool utilizes state-of-the-art natural language processing techniques to filter and visualize requirements, enhancing the process of software reuse.

FeaMod feature extraction results interface
Feature and requirement extraction interface

Key Features:

  • Automated Analysis: Processes `.c` files to discover features and requirements effectively.
  • Intelligent Extraction: Employs machine learning models like BERT and Roberta for accurate requirement filtering.
  • Visualization: Offers bar graphs for comparative coherence analysis, enhancing interpretability of results.
  • User-Friendly Interface: Provides an intuitive interface for easy management and visualization of feature models.

Parallel Computing Technique using OpenMP

This repository explores various techniques to optimize matrix multiplication wall time using OpenMP, a shared-memory multiprocessing API. The study compares serial execution, parallel execution with OpenMP, and optimized parallel execution, aiming to enhance computational performance in multi-threaded environments.

OpenMP matrix multiplication performance graph
Optimized parallel-execution performance

Key Techniques:

  • Serial Execution: Executes matrix multiplication sequentially in a uniprocessor system.
  • Parallel Execution with OpenMP: Utilizes OpenMP for multi-threaded matrix multiplication.
  • Optimized Parallel Execution: Improves performance through memory access optimization and compiler flags.

Twitter Healthcare Expert Finder and Analysis Tool

The Twitter Healthcare Expert Finder and Analysis Tool is an innovative system tailored to sift through the digital noise and accurately identify medical professionals on Twitter. This tool is instrumental for users seeking reliable health information, offering a gateway to verified medical expertise. By harnessing the power of advanced data mining techniques and Twitter’s expansive API, the application excels in pinpointing authentic voices in healthcare amidst the vast social media landscape.

Learning curve for the random forest model
Random forest learning curve
Learning curve for the support vector machine model
Support vector machine learning curve

Key Features:

  • Data Extraction: Processes over 10,000 tweets daily using Twitter's robust API.
  • Tweet Classification: Employs advanced NLP techniques for precise categorization.
  • Real-time Analytics: Features an integrated dashboard to monitor and analyze expert activity and engagement.

UAV Autonomy and Health Monitoring

An active research project investigating LLM-assisted parameter mapping, real-time fault detection, and safety-aware deployment for unmanned aerial platforms.

Illustration of connected autonomous vehicles and unmanned aerial systems
Connected autonomy and health-monitoring concept

publications

talks