Basic Study for Transfer Learning for Autonomous Driving in Car Race of Model Car
Shohei Chiba, Hisayuki Sasaoka
Abstract
Reinforcement learning, deep learning, and deep reinforcement learning can effectively acquire action rules for the autonomous motion of objects. However, some researchers have reported that the machine learning process requires a large amount of learning time. Besides, the process needs to consider the similarity of the environment between the training target and the test target. There is no such thing as driving only on a course learned in advance in actual autonomous driving. In this study, we have used a transfer learning algorithm for autonomous drivings for model cars. The training target for acquiring the learning model and the actual driving courses are changed. In this study, we report on the effectiveness of transfer learning using a model car as the basis for a learning model acquired by reinforcement learning.