Can AI-powered traffic enforcement system augment road safety by influencing driver behavior and perception?

Abstract

Artificial Intelligence powered Enforcement Systems (AIES) are deployed across 675 locations in the state of Kerala, India to mitigate traffic violations and accidents. This fully automated enforcement system utilizes cameras powered by solar energy and 4G LTE (Long-Term Evolution) technology detect violations in real time and inform the drivers through different media. A comprehensive three-stage investigation is carried out in the present work to assess the effectiveness of AIES, which includes a machine learning-driven investigation of driver perception and behavior, crash data analysis, and spatial and temporal analysis of violation behavior. Artificial Neural Network model was instrumental in revealing the role of perception and experience towards change in driving behavior. DBSCAN (Density-Based Spatial Clustering of Applications with Noise) was used to categorize drivers into two segments based on their perception and behavior. Segment 1 showed reserved perception of AIES, with limited behavioral changes while segment 2 demonstrated a favorable perception and significant positive behavioral changes. The findings revealed a significant reduction in fatal accidents demonstrating the effectiveness of the system in enhancing road safety. Spatial analysis indicated a reduction in violations, particularly in areas where AI cameras were prominently placed. Temporal analysis showed marked improvements in helmet usage for both riders and pillion riders, highlighting the role of AIES in raising awareness and compliance. The study established the need for rotating the camera locations as the deterrent effect on violations is found diminishing with distance from the AIES units. The findings suggest that improving the communication and transparency of the system, while carefully considering the level of information revealed, could further reduce violations.