Self-driving car technology is rapidly advancing, but ensuring safety remains paramount. This necessitates robust risk assessment tools, especially for complex driving environments like those found in Italy. This analysis delves into accident data comparing autonomous vehicles (AVs) and human-driven vehicles (HDVs), highlighting key differences and informing the development of image-based risk analysis tools specifically for the Italian context.
General Accident Trends and Comparisons
Analysis of a comprehensive accident dataset reveals significant differences between AV and HDV accidents. AV accidents predominantly involve other vehicles (80%), with pedestrian involvement notably lower (3%) than in HDV accidents (63% vehicles, 15% pedestrians).
AVs exhibit higher accident rates in work zones, during traffic events, and while performing maneuvers like slowing down or changing lanes. However, ‘proceeding straight’ remains the most common pre-accident movement for both AVs (56%) and HDVs (58%). Distinctively, inattention or poor driving behavior contributes to only 1.8% of AV accidents compared to 19.8% of HDV accidents.
Clear weather conditions prevail in most accidents for both vehicle types, but AV accidents occur more frequently in rainy conditions (11% vs. 5% for HDVs). Rear-end collisions are the most prevalent accident type for both.
Further analysis of rear-end collisions shows HDVs rear-ending AVs more frequently (79%) than the reverse (21%). When AVs rear-end HDVs, they are predominantly in conventional (human-driven) mode (72%), potentially due to slower reaction times compared to the autonomous mode.
A comparison between Advanced Driver-Assistance Systems (ADAS) and fully autonomous driving systems (ADS) reveals differences in accident patterns related to weather, road conditions, and pre-accident movements. ADAS accidents are more frequent in rainy conditions and on wet roads, while ADS accidents show a higher incidence in clear weather.
ADAS, primarily designed for highway use, generally involves higher pre-accident speeds compared to ADS, which are often used in more complex urban settings. A random parameter logit model analysis identified ‘Day of the week’ as a significant random parameter influencing AV accident occurrence.
Road, Environment, and Accident Type Findings
A matched case-control logistic regression model reveals that ADS accidents are less likely to occur in rainy weather compared to HDV accidents. This is attributed to the superior object detection capabilities of AV sensors like RADAR in adverse weather. However, ADS accidents are more likely during dawn/dusk conditions, potentially due to challenges in adapting to rapidly changing light conditions.
AVs demonstrate a lower risk of rear-end and broadside collisions compared to HDVs, likely due to faster reaction times and advanced sensor technology.
Pre-Accident Conditions and Accident Outcomes
While most pre-accident movements by ADS reduce accident probability, turning maneuvers increase the likelihood of an accident compared to HDVs. This highlights the challenges AVs face in complex scenarios requiring advanced situational awareness and decision-making. Turning scenarios present specific complexities related to lane selection, trajectory planning, and dynamic adjustments.
ADS accidents are less frequent in scenarios like proceeding straight, running off the road, and entering traffic lanes, likely due to faster reaction times and automated corrective actions.
Conclusion
This comprehensive analysis of accident data provides crucial insights for developing image-based risk assessment tools for self-driving cars, particularly in the context of Italy’s diverse driving environment. Focusing on specific challenges faced by AVs, such as turning maneuvers and adapting to varying light conditions, will be crucial for enhancing safety and accelerating the adoption of autonomous driving technology in Italy. The findings underscore the importance of developing robust perception and decision-making capabilities for AVs to mitigate risks and ensure safe operation on Italian roads.
