Anticipate accidents before they happen.
When a car begins to lose control, the driver almost never notices it immediately; lateral skidding, for example, occurs before the scare, before the steering wheel corrects, before the brake acts, and in electric and autonomous vehicles this delay is not acceptable.
The problem is that the classic vehicle control models were designed for a predictable world, real roads are not like that, the tires deform, the asphalt changes, the grip suddenly disappears, it was then that the researchers decided to stop choosing between physics or artificial intelligence and decided to combine both.
In practice, the car stops simply reacting to the error and begins to anticipate it.
The new approach called physical AI uses traditional physical models to ensure consistency with reality, while the AI learns the behaviors that physics cannot predict alone, in real time, the car begins to estimate invisible states such as the yaw angle before common sensors manage to detect something incorrect.
The system monitors tire data, lateral forces and environmental variations and begins to estimate critical states that common sensors do not directly measure, such as the lateral yaw angle, one of the most important factors for wind stability, but this advance brings an inevitable consequence.
When machines begin to perceive, foresee and decide for themselves, the real challenge stops being the individual vehicle and becomes the coordination between many of them and that is exactly what they are looking for.
