DLR-Logo -> http://www.dlr.de
DLR Portal Home | Imprint | Privacy Policy | Contact | Deutsch
Fontsize: [-] Text [+]

Reinforcement Learning-Based Path Following Control with Dynamics Randomization for Parametric Uncertainties in Autonomous Driving

Ahmic, Kenan and Ultsch, Johannes and Brembeck, Jonathan and Winter, Christoph (2023) Reinforcement Learning-Based Path Following Control with Dynamics Randomization for Parametric Uncertainties in Autonomous Driving. Applied Sciences. Multidisciplinary Digital Publishing Institute (MDPI). doi: 10.3390/app13063456. ISSN 2076-3417.

[img] PDF - Published version

Official URL: https://www.mdpi.com/2076-3417/13/6/3456


Reinforcement learning-based controllers for safety-critical applications, such as autonomous driving, are typically trained in simulation, where a vehicle model is provided during the learning process. However, an inaccurate parameterization of the vehicle model used for train-ing heavily influences the performance of the reinforcement learning agent during execution. This inaccuracy is either caused by changes due to environmental influences or by falsely esti-mated vehicle parameters. In this work, we present our approach of combining dynamics ran-domization with reinforcement learning to overcome this issue for a path-following control task of an autonomous and over-actuated robotic vehicle. We train three independent agents, where each agent experiences randomization for a different vehicle dynamics parameter, i.e., the mass, the yaw inertia, and the road-tire friction. We randomize the parameters uniformly within pre-defined ranges to enable the agents to learn an equally robust control behavior for all possible parameter values. Finally, in a simulation study, we compare the performance of the agents trained with dynamics randomization to the performance of an agent trained with the nominal parameter values. Simulation results demonstrate that the former agents obtain a higher level of robustness against model uncertainties and varying environmental conditions than the latter agent trained with nominal vehicle parameter values.

Item URL in elib:https://elib.dlr.de/194223/
Document Type:Article
Title:Reinforcement Learning-Based Path Following Control with Dynamics Randomization for Parametric Uncertainties in Autonomous Driving
AuthorsInstitution or Email of AuthorsAuthor's ORCID iDORCID Put Code
Ahmic, KenanUNSPECIFIEDhttps://orcid.org/0000-0002-2266-1067UNSPECIFIED
Ultsch, JohannesUNSPECIFIEDhttps://orcid.org/0000-0001-6483-8468UNSPECIFIED
Brembeck, JonathanUNSPECIFIEDhttps://orcid.org/0000-0002-7671-5251UNSPECIFIED
Winter, ChristophUNSPECIFIEDhttps://orcid.org/0000-0002-6949-303XUNSPECIFIED
Date:9 March 2023
Journal or Publication Title:Applied Sciences
Refereed publication:Yes
Open Access:Yes
Gold Open Access:Yes
In ISI Web of Science:Yes
Publisher:Multidisciplinary Digital Publishing Institute (MDPI)
Series Name:Special Issue Technology Development of Autonomous Vehicles
Keywords:reinforcement learning; deep neural networks; dynamics randomization; autonomous driving; motion control; path following control; uncertainty modeling
HGF - Research field:Aeronautics, Space and Transport
HGF - Program:Space
HGF - Program Themes:Robotics
DLR - Research area:Raumfahrt
DLR - Program:R RO - Robotics
DLR - Research theme (Project):R - Intelligent Mobility (SR) [RO]
Location: Oberpfaffenhofen
Institutes and Institutions:Institute of System Dynamics and Control > Vehicle System Dynamics
Deposited By: Ahmic, Kenan
Deposited On:03 Apr 2023 11:13
Last Modified:28 Nov 2023 12:54

Repository Staff Only: item control page

Help & Contact
electronic library is running on EPrints 3.3.12
Website and database design: Copyright © German Aerospace Center (DLR). All rights reserved.