St Germain en Laye, February 7th 2025.
In the near future, autonomous vehicles will encounter significant challenges from dynamic occluded objects that may obstruct their paths. These objects can include pedestrians, as well as a range of vehicles, such as bicycles, cars, and other modes of transportation. Traditional perception tools like cameras, radar, and lidar may not be effective when the road configuration is hostile.
One hypothetical solution involves placing cameras in strategic locations, such as behind hills or in front of buses, to gather information and transmit it to vehicles via V2X (Vehicle-to-Everything) communication. However, this solution is not only costly and difficult to implement, but it is not car manufacturers business.
As a result, stakeholders in the autonomous vehicle industry must focus on detecting and recognizing scenarios that could lead to hidden or occluded objects. This raises important questions: How many additional scenarios will need to be managed? How many millions of miles must be tested? Which particular cases should be considered?
Some teams are exploring the use of probabilities. For example, one might assume that at 2:00 AM, there is unlikely to be anyone hiding behind a curve or a hill—unless it’s one in a million. This mentality leads to accidents; people often say things like, “I didn’t see him; he came out of nowhere” or “Usually, no one comes from there; this is the first time in my life…” Relying on probabilities can be dangerous if safety is a priority.
This is where Nexyad has developed a prudence driving metric. After researching this topic for approximately fifteen years through collaborative projects with experts from 19 countries in road safety, professional drivers, traffic police, and car insurance representatives, they have created a valuable tool.
Nexyad’s software, SafetyNex, measures driving behavior in real-time at a rate of twenty times per second. It can be applied to everything from basic Advanced Driver Assistance Systems (ADAS) to fully autonomous vehicles. The system is founded on extensive knowledge that has been compiled and analyzed over decades.
SafetyNex can work with minimal information in low-cost vehicles or utilize all the data available in a vehicle. This includes its position, trajectory, instantaneous speed, static environment (the geometry of the road, curves, and intersections), signaling attributes, and dynamic environment (other road users, weather conditions, asphalt quality, and temporary hazards such as roadworks or accidents).
Less equipment means more prudence. If certain objects cannot be detected due to a lack of appropriate perception tools, it is essential to adopt a more prudent driving approach. However, even with the most advanced technologies onboard, dealing with dynamic occluded objects remains a challenge. The only solution is a knowledge-based AI that computes driving prudence, ensuring both safety and efficiency on the road: SafetyNex by Nexyad.
#AutonomousVehicles #DrivingSafety #AIinDriving #RoadSafety #OccludedObjects #VehicleTechnology #SmartTransportation #Nexyad #SafetyNex #InnovationInMobility #DrivingPrudenceMetric