Sim-to-Real Reinforcement Learning for Autonomous Driving Using Pseudosegmentation Labeling and Dynamic Calibration

Applying reinforcement learning algorithms to autonomous driving is difficult because of mismatches between the simulation in which the algorithm was trained and the real world. To address this problem, data from global navigation satellite systems and inertial navigation systems (GNSS/INS) were use...

Full description

Saved in:
Bibliographic Details
Main Authors: Jiseong Heo, Hyoung woo Lim
Format: Article
Language:English
Published: Wiley 2022-01-01
Series:Journal of Robotics
Online Access:http://dx.doi.org/10.1155/2022/9916292
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849308890217316352
author Jiseong Heo
Hyoung woo Lim
author_facet Jiseong Heo
Hyoung woo Lim
author_sort Jiseong Heo
collection DOAJ
description Applying reinforcement learning algorithms to autonomous driving is difficult because of mismatches between the simulation in which the algorithm was trained and the real world. To address this problem, data from global navigation satellite systems and inertial navigation systems (GNSS/INS) were used to gather pseudolabels for semantic segmentation. A very simple dynamics model was used as a simulator, and dynamic parameters were obtained from the linear regression of manual driving records. Segmentation and a dynamic calibration method were found to be effective in easing the transition from a simulation to the real world. Pseudosegmentation labels are found to be more suitable for reinforcement learning models. We conducted tests on the efficacy of our proposed method, and a vehicle using the proposed system successfully drove on an unpaved track for approximately 1.8 km at an average speed of 26.57 km/h without incident.
format Article
id doaj-art-125eae5bfdad43e39b4389dbb64bf004
institution Kabale University
issn 1687-9619
language English
publishDate 2022-01-01
publisher Wiley
record_format Article
series Journal of Robotics
spelling doaj-art-125eae5bfdad43e39b4389dbb64bf0042025-08-20T03:54:20ZengWileyJournal of Robotics1687-96192022-01-01202210.1155/2022/9916292Sim-to-Real Reinforcement Learning for Autonomous Driving Using Pseudosegmentation Labeling and Dynamic CalibrationJiseong Heo0Hyoung woo Lim1Agency for Defense DevelopmentAgency for Defense DevelopmentApplying reinforcement learning algorithms to autonomous driving is difficult because of mismatches between the simulation in which the algorithm was trained and the real world. To address this problem, data from global navigation satellite systems and inertial navigation systems (GNSS/INS) were used to gather pseudolabels for semantic segmentation. A very simple dynamics model was used as a simulator, and dynamic parameters were obtained from the linear regression of manual driving records. Segmentation and a dynamic calibration method were found to be effective in easing the transition from a simulation to the real world. Pseudosegmentation labels are found to be more suitable for reinforcement learning models. We conducted tests on the efficacy of our proposed method, and a vehicle using the proposed system successfully drove on an unpaved track for approximately 1.8 km at an average speed of 26.57 km/h without incident.http://dx.doi.org/10.1155/2022/9916292
spellingShingle Jiseong Heo
Hyoung woo Lim
Sim-to-Real Reinforcement Learning for Autonomous Driving Using Pseudosegmentation Labeling and Dynamic Calibration
Journal of Robotics
title Sim-to-Real Reinforcement Learning for Autonomous Driving Using Pseudosegmentation Labeling and Dynamic Calibration
title_full Sim-to-Real Reinforcement Learning for Autonomous Driving Using Pseudosegmentation Labeling and Dynamic Calibration
title_fullStr Sim-to-Real Reinforcement Learning for Autonomous Driving Using Pseudosegmentation Labeling and Dynamic Calibration
title_full_unstemmed Sim-to-Real Reinforcement Learning for Autonomous Driving Using Pseudosegmentation Labeling and Dynamic Calibration
title_short Sim-to-Real Reinforcement Learning for Autonomous Driving Using Pseudosegmentation Labeling and Dynamic Calibration
title_sort sim to real reinforcement learning for autonomous driving using pseudosegmentation labeling and dynamic calibration
url http://dx.doi.org/10.1155/2022/9916292
work_keys_str_mv AT jiseongheo simtorealreinforcementlearningforautonomousdrivingusingpseudosegmentationlabelinganddynamiccalibration
AT hyoungwoolim simtorealreinforcementlearningforautonomousdrivingusingpseudosegmentationlabelinganddynamiccalibration