Politecnico di Torino - Corso Duca degli Abruzzi, 24 - 10129 Torino, ITALY

+39 011 090 6100 info@tech-share.it

Robust Neural Estimation of the Lateral Speed of Vehicles

Lateral velocity estimationneural observerstructured neural network


Method for estimating the lateral speed of land vehicles, using new observers based on neural networks, with architectures designed starting from physical principles. The generalization of the principles of the kinematics operated by the new neural networks allows estimating the lateral speed with high accuracy and guarantees a good robustness of the estimate as the type of vehicle or the operating/environmental condition varies.

Technical features

Lateral velocity estimation plays an essential role in advanced ground vehicle control and stability systems and for trajectory planning in autonomous driving systems. The present invention relates to a lateral speed estimator based on an innovative neural network, which is designed with a structure conforming to the principles of kinematics. The new neural network estimates the lateral velocity with high accuracy by generalizing kinematic equations and can be trained with little data. The kinematic structure on which it is based improves the robustness of the lateral speed estimate, ensuring adequate accuracy even when the type of vehicle or the operating/environmental condition varies. The invention uses the measurement of yaw rate, longitudinal and lateral acceleration, and longitudinal velocity. The developed technology has a TRL 8.

Possible Applications

  • Lateral speed estimation for land vehicles, including high-end sports vehicles;
  • Active support for driving land vehicles, even complex vehicles with many axles;
  • Active support for autonomous driving land vehicles, even complex vehicles with many axles.


  • More accurate estimation of lateral speed, compared with conventional kinematic observers;
  • More robust estimation of lateral speed as vehicle type or the operating/environmental condition varies;
  • Does not require an Inertial Measurement Unit (IMU);
  • Lower sensitivity to noise on measurements.