KIT | KIT-Bibliothek | Impressum | Datenschutz

Actor-Critic Reinforcement Learning for Linear Longitudinal Output Control of a Road Vehicle

Puccetti, Luca ORCID iD icon 1; Rathgeber, Christian; Hohmann, Sören 2
1 Karlsruher Institut für Technologie (KIT)
2 Institut für Regelungs- und Steuerungssysteme (IRS), Karlsruher Institut für Technologie (KIT)

Abstract (englisch):

Reinforcement Learning provides a way to learn optimal controllers but struggles when the state is only partially observed. This is the case with the actuator dynamics in vehicle longitudinal control that translate a controller output into acceleration. We propose a structure for approximating the value function that exploits this property. It allows the critic to learn an observer in the form of a finite impulse response filter and reduces the variance in temporal difference learning. We show that our approach learns faster and with higher precision than an agent that ignores the unobserved states. This is still the case even when using a realistic simulation model with disturbances for the longitudinal dynamics of a road vehicle.


Originalveröffentlichung
DOI: 10.1109/ITSC.2019.8917113
Scopus
Zitationen: 5
Dimensions
Zitationen: 5
Zugehörige Institution(en) am KIT Institut für Regelungs- und Steuerungssysteme (IRS)
Publikationstyp Proceedingsbeitrag
Publikationsmonat/-jahr 10.2019
Sprache Englisch
Identifikator ISBN: 978-1-5386-7024-8
KITopen-ID: 1000105835
Erschienen in IEEE Intelligent Transportation Systems Conference, ITSC 2019, Auckland, New Zealand, 27th - 30th October 2019
Verlag Institute of Electrical and Electronics Engineers (IEEE)
Seiten 2907–2913
Nachgewiesen in Scopus
Dimensions
Relationen in KITopen
KIT – Die Forschungsuniversität in der Helmholtz-Gemeinschaft
KITopen Landing Page