A remarkably powerful technology, AI is transforming virtually every industry, not least of which is the automotive sector, particularly for vehicle safety systems. From the evolution of advanced driver assistance systems (ADAS) to the development of fully autonomous vehicles, AI is driving how cars sense and respond to their environment to protect their occupants. This article examines the burgeoning role of AI in modern car safety systems and how it is making our roads safer as well as look at the possibilities of what could happen as AI continues to evolve.
AI’s Role in Enhancing Vehicle Safety
Passive forms of safety, such as seat belts, airbags and crumple zones that absorb kinetic energy in the event of a collision, have been a long-standing feature of standard vehicles. But in the era of AI, ‘safety has become quite proactive’, Majumdar says: before accidents can even happen, a vehicle’s systems are reaching out to avert danger.
Modern protection systems make use of sensors, cameras, radar and complex algorithms to monitor driving conditions and identify hazards. They also take corrective and preemptive measures when needed in order to reduce the risk of accidents. Assistive and preventive are the key terms here: the driving system is complementing the work of the driver or taking over completely to intervene in some aspect of the drive.
Here are some of the key AI-powered technologies currently shaping modern car safety systems:
- Advanced Driver Assistance Systems (ADAS)
Among the main applications of AI that we can find in cars, there are Advanced Driver Assistance Systems or ‘ADAS’. These systems use AI to process data collected by several sensors (for example, cameras, RADAR or LIDAR) and provide feedback to the driver or intervene on his behalf when necessary. ADAS includes several different features, for example:
Lane Departure Warning and Lane-Keeping Assist
Lane departure warning monitors where a vehicle is in its lane using cameras and sensors, and will issue an alert if the car begins to wander over the line without the turn signal being engaged. Lane-keeping assist will sometimes even take the wheel itself to steer the car into the correct lane to help avoid accidents due to a distracted, fatigued or otherwise impaired human driver.
Adaptive Cruise Control (ACC)
While traditional cruise control maintains a steady speed, adaptive cruise control also uses AI to control vehicle speed, slowing and accelerating as traffic requirements dictate. By tracking the pace of surrounding traffic, and automatically altering your carriage speed to maintain a safe following distance until traffic clears, AI-supported ACC helps reduce the likelihood of rear-end collisions – and the resulting car-crash claims – in situations such as stop-and-go traffic.
Automatic Emergency Braking (AEB)
Being able to distinguish between an accident waiting to happen and false positives are among the most important safety-systems that AI works on. (And also among the most challenging.) One basic AI-driven safety feature is automatic emergency braking. AEB systems incorporate sensors and cameras that feed data into the internet of things (IoT), to determine if there is anything in the path of the vehicle: a pedestrian, a car ahead on the road or a squirrel in the middle of the road. When AEB systems detect a potentially hazardous situation where a collision may otherwise occur, AI algorithms activate, assessing data from cameras and other sensors that are constantly scanning the environment and taking in and computing the information in real time to decide whether to intervene and stop the car, or proceed more slowly.
Blind-Spot Monitoring and Collision Warning Systems
For example, blind-spot monitoring systems rely on AI to detect vehicles in areas outside the driver’s field of view so that any alerts can prevent lane-change collisions by spotting other vehicles sooner. And on the same subject, AI could be applied to collision warning systems with similar usefulness, since it can be anticipated that collision warning systems would be able to assess the relative speed and distances of surrounding vehicles. Surprisingly, collision warning systems could warn a driver to take evasive action if they are in danger of hitting another vehicle.
- Pedestrian and Cyclist Detection
Now, AI has started to demonstrate the ability to detect vulnerable road users (VRUs) such as pedestrians and cyclists. Clear and focused AI-controlled cameras and sensors can detect a pedestrian in a crosswalk or a cyclist on a road that might not even be visible to the driver, even at night or in bad weather. The system can warn the driver in hazardous situations, or even take control to prevent a collision.
This is important especially in a city street with lots of pedestrian and cyclist traffic: AI will need to react very quickly to these unpredictable objects on the road, and consequently will be much less likely to get into a pedestrian-related accident now, compared with a car being driven by a human driver who would take longer to react.
- Driver Monitoring Systems (DMS)
An equally critical domain where AI is advancing safety is through driver monitoring systems (DMS). Driving is a highly mental task and, without a human being behind the wheel, it is impossible to maintain vigilance and detect the earliest symptoms of driver distraction or impairment. AI-powered cameras and sensors are now starting to detect when a driver’s attention is flagging, whether it is due to drowsiness, distraction or intoxication. State-of-the-art AI algorithms analyse micro expressions and eye movements, as well as other physiological indicators such as heart rate to detect minute yet meaningful signals.
If the system detects that the driver is nodding off or is becoming inattentive, it can issue warnings or advise that a stop should be made. The most advanced systems might even take corrective action, such as, slowing down or pulling over to the side of the road and stopping.
- AI in Autonomous Driving
The ultimate promise of AI in the automotive safety setting is the fully autonomous vehicle, or self-driving car. Fully autonomous vehicles have not yet entered the regular market, but significant AI is involved in the systems for autonomous driving.
Self-driving cars rely on AI to:
Process many sources of data, instantaneously, from cameras, radar and LIDAR (a laser detection and ranging system), as well as regular vehicular sensors such as brakes, accelerator and steering wheel.
Recognize and classify objects such as other vehicles, pedestrians, traffic signals, and road signs.
Predict what other road users are going to do next, whether a pedestrian will walk into the road or another vehicle will make an edge case manoeuvre.
Do you drive it? Or does it make decisions for itself based on data gathered by sensors that indicate when it should change lanes, slow down or speed up, so that you, and everyone else on the road, won’t be harmed? Alex Wiltschko and Damien Roch from the Karlsruhe Institute of Technology put it in a 2021 review in the journal Artificial Intelligence Reviews like this: ‘In most cases, the implementation of instructions to the car lies in the hands of AI algorithms that must make decisions for the car … in [athousandths, trillionths, or quadrillionths of a second] in a way that is much better than humans do.’
Granted, full autonomy for vehicular travel has yet to become standard, but partial autonomy already exists via systems such as Tesla’s Autopilot and GM’s Super Cruise. When used as intended, they allow drivers to relax, take their hands off the wheel, and enjoy hands-free highway travel under certain conditions. It seems to me that, as AI becomes even more sophisticated, autonomous vehicle driving on the road will likely be even safer than it is now, and thereby increasingly reliable, while eliminating any number of instances of human error that currently contribute to so many traffic accidents.
- Real-Time Traffic and Hazard Updates
AI is improving individual vehicle safety, too, but also making the driving environment safer overall. AI-based guidance systems such as Daimler’s Mercedes-Benz Navigation, Google Maps and Waze come with real-time traffic information that warns of accidents, road construction or closures, or other hazards, such as cars parked on the shoulder or debris in the road.
These systems use information from other drivers, traffic cameras and city infrastructure before guiding human drivers to safer routes, using AI to interpret that information very quickly to advise of oncoming accidents or congestion, suggest alternative routes, and help drivers avoid any potential hazards while enjoying a smoother overall experience.
- Vehicle-to-Everything (V2X) Communication
AI (Artificial Intelligence) is also used in Vehicle-to-Everything (V2X) communication, which stands for all vehicles (V2V), in conjunction with infrastructure such as traffic lights and road signs (V2I), and with pedestrians and cyclists (V2P).
V2X communication systems use the AI-interpreted data provided from these sources to inform the driver in real-time about upcoming hazards or changes to traffic conditions nearby. A vehicle near an intersection could be given data from a traffic light to provide the vehicle with knowledge about when the light will turn or when a nearby pedestrian will cross the street. If it works as predicted, these systems will reduce accidents and improve road safety to unprecedented degrees.
The Future of AI in Car Safety Systems
But vehicle safety is already one of AI’s fastest-growing fields, and as technology evolves, there’s little doubting that AI’s role in the dashboard is about to grow even more robust – in the near future, the car of the future might have the following new features to offer:
Full automation: Fully autonomous cars that could one day operate in dense urban settings and cross-country without human intervention.
Improved predictive analytics: AI systems will be better at predicting what could go wrong before it actually does, for example anticipating the actions of other drivers or spotting road conditions that could lead to an accident.
Connected car ecosystems: Using artificial intelligence, cars will become part of a network that shares data not only with each other, but also with infrastructure, law enforcement and emergency services, to form a safer, more efficient transportation ecosystem.
Conclusion
The emergence of AI-powered safety systems in today’s automobile is shaping the future of automobiles, and car safety in particular, in ways that were unthinkable even three decades ago. Today’s advanced driver assistance systems (ADAS) with pre-collision systems, electronic stability control, automatic emergency braking, lane-keeping assist, adaptive cruise control, blind spot detection, pedestrian, and cyclist monitoring, and park assist steering; the commercialisation of level 2 driver-assist vehicles and level 3 autonomous vehicles that govern speed and distance of a vehicle from the front and can, under certain conditions, operate without driver supervision; and the development of self-driving ‘autonomous vehicles’ or ‘driverless cars’ systems that take complete control to the car’s wheels are all reducing the chances of human error, enabling timely responses from the vehicle to environmental events even in real-time, and detecting accident-related accidents before they occur. As AI advanced vehicle safety systems continue to mature, the future of car safety looks bright – safer roads, reduced accidents, and a more efficient transportation system for all.
