Graduation Semester and Year
Spring 2024
Language
English
Document Type
Thesis
Degree Name
Master of Science in Computer Science
Department
Computer Science and Engineering
First Advisor
Ming Li
Second Advisor
David Levine
Third Advisor
Abhishek Santra
Abstract
Cycling presents a compelling solution for promoting personal health and environmental well-being, particularly for short-distance travel. Despite its numerous advantages, cycling uptake in the United States remains disproportionately low, primarily due to safety concerns. Traditional frameworks for assessing cyclist stress are hindered by their impracticality and inability to provide real-time evaluations. Self-report surveys and physiological measurements offer alternative approaches but suffer from limitations such as retrospective reporting biases and accessibility challenges, respectively. This thesis introduces CyclistAI, a novel smartphone-based cyclist stress assessment model that leverages context sensing. By combining Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) techniques, CyclistAI aims to provide real-time stress assessment for cyclists. Challenges in dataset annotation due to safety concerns are addressed through simulation-based data generation. Subsequent Domain Adaptation using Contrastive Learning bridges the simulation-to-reality gap and ensures the model’s efficacy in real-world scenarios. In-field testing of CyclistAI demonstrates promising results, showcasing its potential as a pioneering stress assessment tool. Furthermore, this thesis proposes utilizing aggregated assessment results to create stress distribution maps, facilitating informed decision-making for urban planners to enhance cycling infrastructure and promote safer, more sustainable transportation environments.
Keywords
CyclistAI, Cyclist stress assessment, Cross domain contrastive learning, Deep learning, Domain adaptation, Cycling, Smartphone solution, Stress distribution heatmaps
Disciplines
Artificial Intelligence and Robotics | Software Engineering
License
This work is licensed under a Creative Commons Attribution-NonCommercial-Share Alike 4.0 International License.
Recommended Citation
Singh, Aairish, "CyclistAI: A SMARTPHONE SOLUTION FOR CYCLIST STRESS ASSESSMENT USING DEEP LEARNING" (2024). Computer Science and Engineering Theses. 1.
https://mavmatrix.uta.edu/cse_theses/1