Predictive Collision Risk Modeling in Autonomous Driving Using Random Forest and IoT Data

Abstract

The safety of autonomous cars is largely contingent upon their capacity to anticipate and evade possible accidents in real time.  The integration of Internet Things (IoT)-enabled sensors and communication systems allows cars to collect extensive environmental and operational data, facilitating informed decision-making.  This paper presents a predicted collision risk model based on Random Forest (RF) that employs multi-sensor IoT data, including vehicle speed, acceleration, braking habits, lane position, meteorological conditions, and nearby traffic density.  RF is chosen for its resilience, clarity, and capacity to manage high-dimensional, diverse datasets without succumbing to overfitting.  The program evaluates dynamic driving situations to predict accident risks and facilitates preventive actions like controlled braking or lane adjustments.  Experimental findings demonstrate increased prediction accuracy with little computational lag, making it appropriate for real-time implementation.  The proposed approach augments Vehicle-to-Vehicle (V2V) and Vehicle-to-Infrastructure (V2I) awareness, hence enhancing safety and efficiency in autonomous driving systems.