รหัสดีโอไอ 10.14457/TU.the.2019.749
Title Reinforcement learning techniques for identifying social space model of human-robot interaction
Creator Pakpoom Patompak
Contributor Itthisek Nilkhamhang, Advisor
Publisher Thammasat University
Publication Year 2019
Keyword Social robotics ,Human-robot interaction ,Social conventions ,Human robot symbiosis ,Social space ,Machine learning ,Reinforcement learning
Abstract Human-robot interaction in a shared environment is a critical component of service robotics. The ability to perceive, understand, and act in a manner that conforms to a social convention is the fundamental key to human-robot symbiosis. Notably, for navigation tasks, the robot should take into consideration human’s social space, defined as the area that we feel comfortable to interact with other humans or robots. The primary task of the robot is to understand and identify this social space by learning and adjusting to human response. Reinforcement learning, which is a machine learning technique that attempts to maximize the accumulated reward through trial-and-error, can be used to update the parameters of the social space model by learning from previous human-robot interactions. However, different reinforcement learning algorithms exist that may or may not be appropriate for human-robot interaction in a real-world scenario.This thesis focuses on studying the efficacy of reinforcement learning algorithms for parameter adaptation of human’s social space model. We study and analyze the performance of popular reinforcement learning algorithms in terms of respective advantages and disadvantages. Methods for accelerating reinforcement learning to meet real-world requirements are explored. Simulation results are presented and compared that examines the efficiency of each reinforcement learning algorithm, as well as its suitability for adapting the parameters of the social space model.
ดิจิตอลไฟล์ Digital File #1

บรรณานุกรม

Pakpoom Patompak และผู้แต่งคนอื่นๆ. (2019) Reinforcement learning techniques for identifying social space model of human-robot interaction. Thammasat University:ม.ป.ท.
Pakpoom Patompak และผู้แต่งคนอื่นๆ. 2019. Reinforcement learning techniques for identifying social space model of human-robot interaction. ม.ป.ท.:Thammasat University;
Pakpoom Patompak และผู้แต่งคนอื่นๆ. Reinforcement learning techniques for identifying social space model of human-robot interaction. ม.ป.ท.:Thammasat University, 2019. Print.