MENTAL STATES & MACHINE ENHANCING DRIVER ENGAGEMENT IN AUTOMATED VEHICLES FOR SAFER TRANSITIONS
PROJECT 1
PROJECT DESCRIPTION
Automated vehicles still need human drivers to take over when the system reaches limitations. This proposed research aims to conduct three phases of studies to understand the impact of various mental states on takeover performance, develop machine learning models for predicting these interactions, and design an advanced Human-Machine Interface (HMI) tool. Together, these phases seek to enhance traffic operations, bolster infrastructure resilience, and fortify transportation cybersecurity, ensuring a safer and more efficient future for automated driving.
PRINCIPAL INVESTIGATOR(s)
Gaojian Huang
Egbe-Etu Etu
FUNDING SOURCE & AMOUNT
$99,999 Federal, $50,000 Non-Federal Match
PROJECT START DATE
Official Start Date: June 1, 2023
Expected Start Date: November 1, 2023
RESEARCH OUTPUT & IMPACTS
From this project, we anticipate a deeper understanding of how mental states influence automated vehicle takeover performance. We will obtain empirical data from this project, which can be used to model these interactions using machine learning, paving the way for real-time applications of mental state detection methods. Possible findings of the project will 1) provide an HMI tool in alerting and assisting drivers throughout the complex takeover process, including relocating attention, perceiving and processing information, and making and executing decisions; 2) contribute to the literature in developing more substantial frameworks that describe how human drivers interact with AVs; and 3) help engineers, designers, policymakers, and scientists improve the design of next-generation AVs.
Addressing the critical interplay between human mental states and AV takeover, our findings aim to bolster safety by aiding drivers during complex takeovers. By offering a refined HMI tool and contributing to the broader understanding of human-AV interactions, we ensure equitable AV experiences for all users (including users in a wide range of demographic groups, such as older adults, people with disabilities, etc.). Vehicle manufacturers, designers, and policymakers stand to benefit most from our insights, guiding the development and regulation of next-generation AVs. Collaborations with industry stakeholders and outreach activities will further ensure the practical application of our research
PROJECT 2
BUILDING AI AND MACHINE LEARNING TECHNOLOGIES FOR ENHANCING TRANSPORTATION STATION AREA SAFETY IN SAN JOSE, CA
PROJECT DESCRIPTION
Criminal activities often cluster around transportation hubs like transit stations. While accurate crime prediction tools enhance crime prevention, mobility, and transportation equity, few research integrates historical crime data, transportation networks, and hub locations using AI. Our project develops a machine learning algorithm with the implementation of the software package in Python, leveraging a multi-layered geo-statistical model to predict crimes within transportation systems, enhancing safety and increasing ridership. Unlike past tools focusing solely on historical crime or land use data, our tool combines transportation network insights with hub locations taking advantage of rigorous statistical model. This software package enables local jurisdictions to allocate resources more effectively, plan interventions, and strengthen public safety.
PRINCIPAL INVESTIGATOR(s)
Dr. Bo Yang (PI)
Dr. Ahoura Zandiatashbar (Co-PI)
Department of Urban and Regional Planning, San Jose State University
FUNDING SOURCE & AMOUNT
$100,043 Federal, $50,021 Non-Federal Match
PROJECT START DATE
NOVEMBER 1, 2023
RESEARCH OUTPUT & IMPACTS
In this project, our goal is to develop an AI-driven crime prediction tool using a comprehensive spatio-temporal geo-statistical model, implemented with Python and grounded in the latest Spatio-temporal Cokriging Method. This tool is designed to anticipate criminal activities within transportation systems, enhancing safety and increasing ridership. Our method harnesses both historical crime data and transportation network insights. Real-time predictions will empower transportation authorities to strategize, allocate resources, and preempt potential criminal activities, leading to anticipated decreases in crime rates and bolstering public trust in transportation systems. The proposed tool seamlessly integrates Python with ArcGIS and Google Earth Engine, providing an intuitive platform for spatial analysis. It offers a user interface for inputting data and visualizing real-time crime predictions and hotspots, revolutionizing transportation safety measures. Authorities will be equipped to pinpoint and reinforce high-risk zones promptly. This tool will be smoothly incorporated into existing surveillance systems at transportation hubs, optimizing security protocols and ensuring efficient law enforcement resource deployment.
Our proposed crime prediction tool integrates Python with ArcGIS Online and Google Earth Engine, creating an intuitive platform for spatial analysis. The user-friendly interface facilitates data input and displays crime predictions and live hotspots, heralding a transformative approach to transportation safety. Authorities gain real-time insights into potential high-risk areas, enabling proactive risk management. This model bolsters security in critical zones, ensuring passenger safety and optimizing law enforcement resource deployment. The tool will be integrated seamlessly into existing surveillance infrastructure at transportation centers. Its broader impact is profound. By discerning recurring crime patterns, transport planners and urban designers can embed innovative security measures directly into transportation hub designs—be it strategic surveillance, improved lighting, or other safety features. Insights will inform policy, advocating for standardized safety designs in future transportation projects. The system's strength lies in its shift from merely reactive to predominantly proactive safety protocols.