The loss of life from road traffic accidents is estimated to be around 1.35 million people annually according to data from the World Health Organisation in 2020. It is further estimated that no fewer than 20 million people suffer non-fatal injuries from accidents. A significant percentage of those affected include pedestrians, bicyclists and motorcyclists. These figures are alarming, prompting response from the United Nations General Assembly to prevent at least 50% of road accidents by 2030.
One step towards achieving this aim is acknowledging the reality of the statistic that around 90% of accidents are due to human error which could be of different kinds - overspeeding, text driving, inattention, etc. Making mistakes - whether deliberate or not - is written into human DNA. Several safety measures have been adopted in vehicular transport, such as seatbelts, airbags, traffic regulation, etc. These safety measures can be advanced through the application of Artificial Intelligence.
Introducing AI into road safety can contribute to achieving the UN’s goals as it encourages the adoption of proactive measures that seek to diagnose the inherent risks in road networks and driving and solving these challenges before they create more danger to road users.
Artificial intelligence to Improve Road Safety
While fully autonomous self-driving vehicles may not be in mainstream use yet, AI can still augment human drivers and improve safety in a number of ways. In broad terms, AI techniques such as computer vision can be applied with sensors and other such systems to gather data on road infrastructure and traffic to draw valuable insight from the detected patterns. Furthermore, predictive AI models can be used for risk stratification - to identify road areas with high risk of accidents. This simplified explanation doesn’t capture the sheer amount of complexity that AI will tackle to achieve this. However, the technology is capable of pattern detection and insight identification in never-before imagined accuracy and speed.
Some of the ways by which AI can improve road safety are discussed as follows:
i. Risk stratification: AI can be used to analyse video log data around road networks to identify possible risk factors such as obstacles, steep slopes, enbankments, etc. This can be used to stratify road networks according to the risk of accident occurence. The government can use this data to prioritise resource allocation to tackle this issue. Existing data on accidents can also be used to develop a better understanding of unsafe road networks and measures to introduce safety.
ii. Collision avoidance: Through computer vision, cars can learn to identify obstacles - especially when they appear suddenly. This quick recognition can be helpful in situations where intuitive decision-making is required. Based on pre-configured actions, semi-autonomous vehicles may make the decision to apply brakes or swerve. This application can be very useful in the case of lane cut-ins too.
iii. Computer Vision-powered Parking Management: AI's capacity to process vast amounts of data input can be combined with Computer Vision to create a smart parking network management system that is updated in real time and feeds road users with information on available parking spots. This can help to lessen traffic congestion problems which in its own way can improve road safety.
iv. Driver behavior monitoring: Through the use of dashboard cameras and facial recognition technology, high-risk behavior (e.g. drowsiness) in a driver can be identified and trigger alarms to prevent any possible crash. The technology can also be used to identify other forms of risky behavior such as distraction from the road watching.
v. Automated license plate recognition: Road safety can be improved in unsafe drivers are taken off the road. Therefore, automated license plate recognition software can be used to process multiple plates at once and cross-reference them with a database of drivers with records.
Postl, Raphael. (2021). Artificial intelligence: Paving the Way for Road Safety. 10.13140/RG.2.2.28450.15047.