Control longitudinal de un vehículo mediante aprendizaje por refuerzo profundo
- Barreno Herrera, Felipe 1
- Santos, Matilde 1
- Romana, Manuel 1
-
1
Universidad Complutense de Madrid
info
- Ramón Costa Castelló (coord.)
- Manuel Gil Ortega (coord.)
- Óscar Reinoso García (coord.)
- Luis Enrique Montano Gella (coord.)
- Carlos Vilas Fernández (coord.)
- Elisabet Estévez Estévez (coord.)
- Eduardo Rocón de Lima (coord.)
- David Muñoz de la Peña Sequedo (coord.)
- José Manuel Andújar Márquez (coord.)
- Luis Payá Castelló (coord.)
- Alejandro Mosteo Chagoyen (coord.)
- Raúl Marín Prades (coord.)
- Vanesa Loureiro-Vázquez (coord.)
- Pedro Jesús Cabrera Santana (coord.)
Verlag: Servizo de Publicacións ; Universidade da Coruña
ISBN: 9788497498609
Datum der Publikation: 2023
Seiten: 127-131
Kongress: Jornadas de Automática (44. 2023. Zaragoza)
Art: Konferenz-Beitrag
Zusammenfassung
This paper presents an intelligent system for longitudinal control of a vehicle using deep reinforcement learning based on the influence of road curvature. The intelligent system consists of an agent based on the deep deterministic policy gradient (DDPG) algorithm for speed control. To train the model agent, the road curvature effect is considered through the perceived acceleration obtained from the lateral acceleration and angular velocity due to the road itself. The results of the intelligent system are continuous acceleration values. The proposed model offers promising results, suggesting that this intelligent system can assist the driver and that the vehicle control system can be applied to semi-autonomous or autonomous driving making driving safer and more comfortable.