2518-1092Research result. Information technologies2518-109210.18413/2518-1092-2024-9-2-0-83495ARTIFICIAL INTELLIGENCE AND DECISION MAKING<strong>COMPARATIVE ANALYSIS OF DEEP LEARNING ALGORITHMS WITH REINFORCEMENT DDPG, PPO AND SAC FOR UNMANNED CAR CONTROL IN CARLA SIMULATOR</strong><strong>COMPARATIVE ANALYSIS OF DEEP LEARNING ALGORITHMS WITH REINFORCEMENT DDPG, PPO AND SAC FOR UNMANNED CAR CONTROL IN CARLA SIMULATOR</strong>TikhonovMaksim KonstantinovichTikhonovMaksim Konstantinovichsamualgame@gmail.com20249200

This paper presents a comparative analysis of three advanced deep reinforcement learning algorithms: Deep Deterministic Policy Gradient (DDPG), Proximal Policy Optimization (PPO) and Soft Actor-Critic (SAC) implemented in the Stable Baselines 3 library. The aim of the study is to evaluate the performance and applicability of each of the algorithms for the task of driving an unmanned vehicle in the complex and dynamic environment provided by the CARLA simulator, focusing on key metrics such as total distance, total reward, average speed, deviation from the center of the roadway, and success rate of episodes. The authors describe the experimental testing methodology in detail, including the tuning of training parameters and performance evaluation criteria. Experimental results demonstrate differences in the performance of the algorithms, revealing their strengths and weaknesses in the context of autonomous driving. The paper contributes to the understanding of the advantages and limitations of each algorithm in the context of autonomous driving and offers recommendations for their practical application.

This paper presents a comparative analysis of three advanced deep reinforcement learning algorithms: Deep Deterministic Policy Gradient (DDPG), Proximal Policy Optimization (PPO) and Soft Actor-Critic (SAC) implemented in the Stable Baselines 3 library. The aim of the study is to evaluate the performance and applicability of each of the algorithms for the task of driving an unmanned vehicle in the complex and dynamic environment provided by the CARLA simulator, focusing on key metrics such as total distance, total reward, average speed, deviation from the center of the roadway, and success rate of episodes. The authors describe the experimental testing methodology in detail, including the tuning of training parameters and performance evaluation criteria. Experimental results demonstrate differences in the performance of the algorithms, revealing their strengths and weaknesses in the context of autonomous driving. The paper contributes to the understanding of the advantages and limitations of each algorithm in the context of autonomous driving and offers recommendations for their practical application.

deep reinforcement learningautonomous drivingDDPGPPOSACStable Baselines 3CARLAdeep reinforcement learningautonomous drivingDDPGPPOSACStable Baselines 3CARLAСписок литературыLillicrap T.P. et al. Continuous control with deep reinforcement learning // arXiv preprint arXiv:1509.02971. &ndash; 2015.Chang C.C. et al. Autonomous driving control using the ddpg and rdpg algorithms // Applied Sciences. &ndash; 2021. &ndash; Т. 11. &ndash; №. 22. &ndash; С. 10659.Schulman J. et al. Proximal policy optimization algorithms // arXiv preprint arXiv:1707.06347. &ndash; 2017.Emuna R., Borowsky A., Biess A. Deep reinforcement learning for human-like driving policies in collision avoidance tasks of self-driving cars // arXiv preprint arXiv:2006.04218. &ndash; 2020.Haarnoja T. et al. Soft actor-critic: Off-policy maximum entropy deep reinforcement learning with a stochastic actor // International conference on machine learning. &ndash; PMLR, 2018. &ndash; P. 1861-1870.Ke P., Yanxin Z., Chenkun Y. A decision-making method for Self-driving based on deep reinforcement learning // Journal of Physics: Conference Series. &ndash; IOP Publishing, 2020. &ndash; Т. 1576. &ndash; №. 1. &ndash; P. 012025.Youssef F., Houda B. Comparative study of end-to-end deep learning methods for self-driving car // International Journal of Intelligent Systems and Applications. &ndash; 2020. &ndash; Т. 12. &ndash; P. 15-27.Li D., Okhrin O. Modified DDPG car-following model with a real-world human driving experience with CARLA simulator // Transportation research part C: emerging technologies. &ndash; 2023. &ndash; Т. 147. &ndash; P. 103987.