Summary
International Symposium on Nonlinear Theory and its Applications
2017
Session Number:A1L-B
Session:
Number:A1L-B-5
Control of Nonholonomic Vehicle System Using Hierarchical Deep Reinforcement Learning
Naoyuki Masuda, Toshimitsu Ushio,
pp.26-29
Publication Date:2017/12/4
Online ISSN:2188-5079
DOI:10.34385/proc.29.A1L-B-5
PDF download (183.4KB)
Summary:
In this paper, we apply an approach integrating two reinforcement algorithms to a parking problem of 4-wheeled vehicle and obtain a controller that generates an optimal trajectory. One enables exploring more efficient by hierarchizing a learning agent and the other enables learning from a unshaped reward. By simulation, we show that by hierarchizing the policy of the agent, a parking operation including cutting of the wheel, which requires a long exploration before acquisition of the movement, can be acquired efficiently.