Deep Reinforcement Learning for a Self-Driving Vehicle Operating Solely on Visual Information

被引:2
作者
Audinys, Robertas [1 ]
Slikas, Zygimantas [1 ]
Radkevicius, Justas [1 ]
Sutas, Mantas [1 ]
Ostreika, Armantas [1 ]
机构
[1] Kaunas Univ Technol, Fac Informat, Dept Multimedia Engn, LT-51368 Kaunas, Lithuania
关键词
autonomous driving; Vision Transformers (ViT); Deep Reinforcement Learning (DRL); policy learning; metadrive; airsim; safe exploration; FRAME;
D O I
10.3390/electronics14050825
中图分类号
TP [自动化技术、计算机技术];
学科分类号
0812 ;
摘要
This study investigates the application of Vision Transformers (ViTs) in deep reinforcement learning (DRL) for autonomous driving systems that rely solely on visual input. While convolutional neural networks (CNNs) are widely used for visual processing, they have limitations in capturing global patterns and handling complex driving scenarios. To address these challenges, we developed a ViT-based DRL model and evaluated its performance through extensive training in the MetaDrive simulator and testing in the high-fidelity AirSim simulator. Results show that the ViT-based model significantly outperformed CNN baselines in MetaDrive, achieving nearly seven times the average distance traveled and an 87% increase in average speed. In AirSim, the model exhibited superior adaptability to realistic conditions, maintaining stability and safety in visually complex environments. These findings highlight the potential of ViTs to enhance the robustness and reliability of vision-based autonomous systems, offering a transformative approach to safe exploration in diverse driving scenarios.
引用
收藏
页数:30
相关论文
共 35 条
[1]  
Alijani S, 2024, Arxiv, DOI arXiv:2404.04452
[2]  
Ando A, 2023, Arxiv, DOI arXiv:2301.10222
[3]   Experimenting Agriculture 4.0 with Sensors: A Data Fusion Approach between Remote Sensing, UAVs and Self-Driving Tractors [J].
Barrile, Vincenzo ;
Simonetti, Silvia ;
Citroni, Rocco ;
Fotia, Antonino ;
Bilotta, Giuliana .
SENSORS, 2022, 22 (20)
[4]   Integration of an Innovative Atmospheric Forecasting Simulator and Remote Sensing Data into a Geographical Information System in the Frame of Agriculture 4.0 Concept [J].
Bilotta, Giuliana ;
Genovese, Emanuela ;
Citroni, Rocco ;
Cotroneo, Francesco ;
Meduri, Giuseppe Maria ;
Barrile, Vincenzo .
AGRIENGINEERING, 2023, 5 (03) :1280-1301
[5]  
Bojarski M, 2016, Arxiv, DOI arXiv:1604.07316
[6]   DeepDriving: Learning Affordance for Direct Perception in Autonomous Driving [J].
Chen, Chenyi ;
Seff, Ari ;
Kornhauser, Alain ;
Xiao, Jianxiong .
2015 IEEE INTERNATIONAL CONFERENCE ON COMPUTER VISION (ICCV), 2015, :2722-2730
[7]  
Chen D., 2019, CORL
[8]  
Chen XZ, 2017, Arxiv, DOI [arXiv:1611.07759, 10.48550/ARXIV.1611.07759]
[9]  
Codevilla F, 2018, Arxiv, DOI arXiv:1710.02410
[10]  
Crayton T., 2017, Autonomous Vehicles: Developing a Public Health Research Agenda to Frame the Future of Transportation Policy