Artificial Intelligence-Driven All-Terrain Vehicle Crash Prediction and Prevention System

Area of Science:

• * Public Health and Safety
• * Artificial Intelligence and Machine Learning
• * Transportation Safety

Background:

• * All-Terrain Vehicle (ATV) crashes pose a significant public health risk in the U.S., leading to numerous fatalities and hospitalizations.
• * Existing ATV safety systems primarily focus on post-crash detection, lacking proactive, rider-assistive preventive measures.
• * Rider's ability to assess risks is crucial for preventing incidents, yet practical tools for real-time risk assessment are underdeveloped.

Purpose of the Study:

• * To develop an AI-driven system for real-time prediction and prevention of All-Terrain Vehicle (ATV) crashes.
• * To create a practical application that assists riders in avoiding imminent crash situations.
• * To specifically address the prediction of rollover incidents, a common and dangerous type of ATV crash.

Main Methods:

• * Development of a deep neural network model to estimate real-time crash likelihood.
• * Integration of an embedded system on the ATV to collect ride parameters (speed, roll/pitch angles).
• * Utilization of a smartphone application for inputting ATV characteristics and rider presence.

Main Results:

• * The AI system achieved a rollover prediction accuracy exceeding 99%.
• * The system demonstrated high precision in detecting roll and pitch angles with average errors of 0.26 and 0.54 degrees.
• * Accurate detection of ATV speed was achieved with an average error of 0.75 m/s.

Conclusions:

• * The developed AI system offers a novel approach to ATV safety through predictive rollover incident detection.
• * The system's high accuracy and real-time capabilities have the potential to significantly decrease ATV-related injuries and fatalities.
• * This technology enables preemptive actions by riders, enhancing overall safety and mitigating risks associated with ATV use.