By Nisarg Pandya
Artificial intelligence (AI) is transforming the automotive industry, particularly in enhancing road safety and reducing accidents. In India, where only 2.5 per cent of the world's vehicles contribute to 13 per cent of global road accidents, the need for innovative solutions is pressing. AI-powered technologies are emerging as vital tools to address this challenge, improving both safety and efficiency in transportation.
One of the most significant contributions of AI to road safety is its predictive capabilities. AI systems can analyse data from various sources, including telematics and driver behaviour, to foresee potential accidents. The perception details fundamentally resemble how humans drive using natural intelligence by continuously answering three questions – where am I? What do I see? And how do I proceed?
The first question defines localisation of information about the exact location based on geography and lane position, the second question is the perception i.e. understanding the surroundings and the third question is decision-making about whether to brake or accelerate. AI solutions use localisation and perception to understand the surroundings and driver behaviour data and assert drivers in case of any risk to brake or accelerate. AI solutions with such intelligence use cameras, radar-like sensors and a computer on the edge to make real-time decisions.
Identifying accident-prone areas and contributing to the design of safer road infrastructure
The identification of accident-prone areas is an outcome of historical behavioural data analysis. By deploying telematics devices across regions, AI can collect data on the surrounding environment such as high traffic areas, accident-prone spots with pedestrian or animal crossings, road conditions from potholes to new constructions, that may impact driving. By consistently analysing such data points, over time AI can identify such areas and notify drivers with necessary voice assistance when approaching them.
AI for Safer Driving
AI devices equipped with ADAS and DMS features use voice assistance to alert drivers in their preferred language about unsafe driving practices, such as using a phone while driving or not wearing a seat belt. These alerts are the first level of education for drivers to proactively improve behaviour in real-time to avoid impact. The second level of education is when drivers receive personalised training from their fleet managers by reviewing videos of their driving behaviour and the potential consequences of unsafe driving practices. By consistently engaging in this training for at least six months, drivers can significantly improve their behaviour. They may become more likely, following the real-time guidance provided by ADAS/DMS devices, leading to proactive decisions and a reduction in risky driving situations.
To be effective, ADAS/DMS devices using AI must accurately identify risky areas for drivers. Inaccurate alerts can diminish driver trust in the technology, leading to a potential disregard for warnings even in dangerous situations. This can hinder behavioural improvements.
Challenges & strategies in implementing road safety
The biggest challenge in implementing AI for road safety, particularly with systems like ADAS, is its ability to generalise across diverse geographic regions. For example, for ADAS to function effectively, it must first recognise key regional differences, such as whether the driving system is left-hand or right-hand, the types of vehicles on the road, variations in traffic lights and signs, differences in sidewalks, and lane markers. The AI must be trained to handle these variations and evolve to become sufficiently generalised to detect and respond appropriately, despite such differences.
To implement DMS–driver behaviour monitoring, it must be able to recognise a wide variety of facial features, such as body build, height, tattoos, glasses etc. Additionally, DMS systems must be adaptable to different cultural and geographical factors that can influence facial expressions as well as orientations and AI must continually evolve over all such variations.
The best strategy to overcome all such challenges is to work on two specific performance metrics of AI- accuracy and precision, which means identifying the false positives and correcting them, and identifying the true negatives and improvising. To achieve both parameters with higher values, one needs a large variety of data to consider the variation in cases concerning different contexts.
Future trends and research directions in AI applications for road safety
AI's role in road safety extends from driver assistance systems like ADAS and DMS to systems like L4/L5 autonomy where AI decisions are used for controlling vehicles. While the world is moving towards autonomous driving, the complexity and context-dependent nature of driving makes it a challenging task for AI. Human drivers, with their diverse experiences and decision-making abilities, can handle various situations differently, sometimes leading to accidents or successful navigation. Achieving the same level of precision and adaptability in AI systems remains a significant challenge, limiting the deployment of fully autonomous vehicles to specific areas and controlled environments.
AI can proactively predict the likelihood of accidents by identifying potential risk factors, such as drowsy driving, tailgating, or distracted driving. While sensors can detect accidents after they occur, AI's ability to anticipate potential incidents allows for preventive measures and faster emergency response.
(The author is the CEO of DrivebuddyAI)
Disclaimer: The opinions, beliefs, and views expressed by the various authors and forum participants on this website are personal and do not reflect the opinions, beliefs, and views of ABP Network Pvt. Ltd.