Vehicles capable of acting where humans can hardly operate
Electric cars are no longer just battery-powered machines but are becoming intelligent systems capable of interpreting their own behavior in real time, and a new advance shows that the next stage of autonomous driving may not depend solely on more sophisticated sensors, but on something deeper: understanding the physics of the vehicle while it is in motion.
Researchers led by Professor Kanghyun Nam developed a system that combines traditional physical models with artificial intelligence to estimate critical vehicle states in real time, especially those that cannot be directly measured by sensors.
One of the main challenges is the so-called lateral skid angle, which is what is revealed when the car begins to slide during curves or on surfaces with little grip. Detecting this moment early is essential because any delay in response can compromise braking, stability and control.
Something even more critical when the driver is software.
The problem is that the behavior of tires is constantly changing. Temperature, speed, wear and type of asphalt completely alter the dynamics of the vehicle, making classic models insufficient for real situations. The solution created by the team was a hybrid architecture. Instead of abandoning physics, they reinforced traditional models with data-driven learning.
The result is a system capable of adapting calculations in real time, understanding changes in the behavior of the tires and the environment much more quickly than conventional methods. When machines manage to understand their own movement in real time, they gain something essential, the ability to act in environments where humans can hardly operate and that is exactly what is allowing automation to be taken to extreme places.
