Research Projects
My research develops geometry-informed computer vision and machine learning methods for reliable perception and decision support in safety-critical applications. My dissertation work addresses transportation safety through two complementary tracks: data-driven behavioral modeling and geometry-informed computer vision, providing experience with both paradigms and their limitations.
Current Research
Fine-Grained Vehicle Type Analysis for Cyclist Overtaking Behavior
Current and final phase of doctoral work integrating geometry-informed computer vision components to investigate how fine-grained vehicle types (e.g., SUVs vs. sedans) systematically influence overtaking behavior. This research builds on geometry-based passing distance estimation (MAE 9.5 cm) and overtaking detection systems to provide comprehensive analysis of cyclist-vehicle interactions.
Dissertation Projects
A Geometry-Informed Computer Vision Method for Detecting and Examining Overtaking Vehicles From A Bicycle
Geometry-informed computer vision system for detecting vehicle overtaking events from bicycle-mounted cameras, achieving 98.7% precision by leveraging perspective geometric principles rather than relying on data-driven appearance models. The system provides 2.89 seconds of early warning time and was validated on 319 real-world overtaking events across 41,500 frames.
TRB Annual Meeting (Accepted), 2026
Geometry-Informed Distance Estimation from 2D Bounding Boxes for Vehicle Overtaking Analysis
Geometry-based passing distance estimation method that achieves MAE 9.5 cm using only monocular vision features. This work demonstrates that explicit mathematical models of geometry can provide robust and computationally efficient solutions for safety-critical estimation tasks.
In preparation, 2025
A Comparative Analysis of Acceleration and Deceleration Profiles for Aggressive Driving Styles and Fuel Economy Test Cycles
Comparative analysis of acceleration and deceleration patterns between real-world aggressive driving behaviors and EPA fuel economy test cycles. The study reveals discrepancies between standardized testing protocols and actual driving patterns, with implications for vehicle efficiency ratings and sustainable transportation policy. Published at SAE World Congress Experience (WCX) 2025.
A Machine Learning Pipeline Framework to Identify Aggressive Driving Based on Vehicle Kinematics and Driver's Pedal Operations
Novel speed-adjusted jerk thresholding method to engineer features grounded in vehicle kinematics and driver pedal operational characteristics. The best model identified aggressive driving with 94% accuracy, significantly outperforming standard baselines. Validated across 556 driving trips with rigorous cross-validation.
Under review, 2025
Master's Thesis
Computational Human Performance Modeling using Queuing Network in an Open-Source Platform
Open-source implementation of the QN-MHP (Queuing Network-Model Human Processor) cognitive architecture using Python and SimPy for computational user interface testing. The model simulates human cognition as a queuing network of information processors, enabling UI evaluation without extensive human subject testing. Model validation using visual-manual tasks demonstrated alignment with empirical data.
Graduate Research Projects
Intelligent Lane Change Prediction for Autonomous Vehicles using Temporal Convolutional Networks
Deep learning system for predicting vehicle lane change behaviors using Temporal Convolutional Networks (TCN). Using real-world trajectory data from the NGSIM dataset, the model achieved 98.66% prediction accuracy with rapid computational time, suitable for real-time deployment in autonomous vehicle decision-making systems.
Mixed Methods Study: Data Privacy Perceptions in Messenger Applications
Mixed-methods research exploring user perceptions of data privacy in messaging applications, comparing WhatsApp and Signal. The study analyzed the balance between privacy, trust, and usability, revealing users' preference for transparency and aggregate data usage over personal data collection.
An Autonomous Driving System - Dedicated Vehicle for People with ASD and their Caregivers
Specialized autonomous vehicle (ADS-DV) design tailored for individuals with Autism Spectrum Disorder through human-centered design and stakeholder engagement with caregivers. Developed customized interior design and companion mobile application addressing safety, monitoring, and individual preferences. Published at ACM AutomotiveUI 2021.