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.
논문 기본 정보
- 자료유형
- 학위논문
- 저자정보
- 지도교수
- 박신석
- 발행연도
- 2020
- 저작권
- 고려대학교 논문은 저작권에 의해 보호받습니다.
이용수10
이 논문의 연구 히스토리 (2)
- 이 논문의 후속연구가 궁금하신가요?
- 연관 학술논문 또는 학술발표를 통해 보다 발전된 연구결과를 확인하실 수 있습니다.
- 이 논문의 연구 히스토리 확인하기
초록· 키워드
오류제보하기
Based on people''s driving habits, this study proposes algorithms using reinforcement learning on how to control vehicle steering and speed, where self-driving vehicles are safe and can quickly avoid obstacles. The algorithms proposed in this study use a much more realistic road condition as an environment, and various sensors enable real-time recognition and control. Using the simulator to conduct obstacle avoidance experiments, using the characteristics of those obtained here, set the criteria for learning, and set up actions and rewards to reflect the tendency to drive quickly and safely. Using the bicycle model as an agent, the dynamic elements of the vehicle are reflected in the learning. It was also possible to achieve practical results by making the learning environment resemble real roads. This result was also applicable to the full-vehicle model. The algorithm also demonstrated successful avoidance based on appropriate speed control and steering control for various barrier layouts, confirming that it can be used universally in various obstacle situations. In addition, comparisons with previous experimental data have confirmed that this algorithm behaves similarly to humans within a defined range.
In this study, we established an obstacle avoidance strategy using hardening learning and applied it to actual simulation vehicles to show similar vehicle behavior in obstacle situations. The results above also identified the potential for a driving control strategy that uses a hardening learning algorithm to demonstrate vehicle behavior similar to that of a real person.
In this study, we established an obstacle avoidance strategy using hardening learning and applied it to actual simulation vehicles to show similar vehicle behavior in obstacle situations. The results above also identified the potential for a driving control strategy that uses a hardening learning algorithm to demonstrate vehicle behavior similar to that of a real person.
목차
제 1장 서 론 11.1 연구 배경 11.2 연구 동향 61.3 연구의 목적 및 내용 9제 2장 강화학습 (Reinforcement Learning) 112.1 강화학습 (Reinforcement Learning) 122.2 강화학습 모델 구성 23제 3장 시뮬레이션 환경 263.1 실제 차량에 사용되는 Sensor 263.2 Pygame 293.3 CarMaker 30제 4장 Action / Reward 설정 354.1 Simulator를 사용한 실제 주행 및 결과 364.2 Action 설정 504.3 Reward 설정 54제 5장 학습 및 학습결과 595.1 전체 학습 환경 구성 595.2 장애물 회피 강화학습 및 결과 615.3 알고리즘 검증 69제 6장 결 론 75참고 문헌 77
최근 본 자료
댓글(0)
0