지원사업
학술연구/단체지원/교육 등 연구자 활동을 지속하도록 DBpia가 지원하고 있어요.
커뮤니티
연구자들이 자신의 연구와 전문성을 널리 알리고, 새로운 협력의 기회를 만들 수 있는 네트워킹 공간이에요.
이용수14
국문요약 viii제 1 장 Introduction 11.1 Background 11.2 Research Trends 41.3 Scope 91.4 Functional Architecture 11제 2 장 Reinforcement Learning 122.1 Definition of Reinforcement Learning 122.2 Concept of Reinforcement Learning 142.2.1 Markov Decision Process&Bellman Equation 142.2.2 Dynamic Programing 182.2.3 Monte-Carlo 202.2.4 SARSA&Q-learning 22제 3 장 Environment Construction 273.1 ROS Message Communication 293.2 Roll of Window PC 313.2.1 Road Environment Construction 313.2.2 Prscan&Simulink Interworking 333.3 Roll of Linux PC 393.3.1 State&Action Definition 393.3.2 Reward Design 41제 4 장 Deep-Q-Network Algorithm 464.1 Deep Neural Network 464.2 Common with DeepMind DQN 524.2.1 Experience Replay 524.2.2 Target Network 554.3 Difference with DeepMind DQN 584.3.1 Asynchronous Learning 594.3.2 Signal to Signal 61제 5 장 Simulation 635.1 Obstacle Avoidance Result based on DQN 655.1.1 Repeated Trial Result of 2000 Episodes 665.1.2 Repeated Trial Result of 10000 Episodes 685.2 Comparison DQN with Reference Algorithm 705.2.1 Car’s Trajectory and Steering angle at 15kph 715.2.2 Car’s Trajectory and Steering angle at 30kph 735.2.3 All Together Comparison at 30kph 77제 6 장 Conclusion 80References 82Abstract 84감사의 글 86
0