作者:楊明珠(大連交通大學(xué), 遼寧 大連 116021)
摘要:人工智能和自動(dòng)駕駛的發(fā)展如火如荼,人工智能技術(shù)在自動(dòng)駕駛技術(shù)中的應用也越來(lái)越多、越來(lái)越深入,因此本文提出一種基于深度強化學(xué)習的單目視覺(jué)自動(dòng)駕駛決策系統,可以讓自動(dòng)駕駛車(chē)輛通過(guò)相機作為傳感器輸入,實(shí)現端到端的自動(dòng)駕駛決策,并且通過(guò)設置獎勵函數和訓練使得車(chē)輛的學(xué)習效率越來(lái)越高,可以在車(chē)道內保持直線(xiàn)行駛。
關(guān)鍵詞:深度強化學(xué)習;自動(dòng)駕駛;決策;單目視覺(jué)
Abstract: The development of artificial intelligence and autonomous driving is in full swing, and the application of artificial intelligence technology in autonomous driving technology is more and more indepth,so this paper proposes a monocular driving auto-driving decision system based on deep reinforcement learning, which can let the selfdriving vehicle use the camera as a sensor input to achieve end-to-end autonomous driving decisions, and set the reward function and training to make the vehicle's learning efficiency more and more efficient, and it can keep driving straight in the lane.
Key words: Deep reinforcement learning; Autonomous driving; Decision making; Monocular vision
在線(xiàn)預覽:基于深度強化學(xué)習的單目視覺(jué)自動(dòng)駕駛決策系統
摘自《自動(dòng)化博覽》2020年5月刊