Autonomous Behavior of Bipedal Robot by Learning Time-series Camera Images

Abstract: The author is conducting basic research on the autonomous behavior of a small biped robot. The system under study acquires behavioral data when a human controls the small biped robot. This system then learns from this behavioral data and image data obtained from the robot’s onboard camera. However, our previous method did not account for time-series behaviors, resulting in the repetition of certain behaviors. To address this issue, this paper utilizes Recurrent Neural Network (RNN), which are well-suited for learning time-series information. As a result, it was confirmed that the robot could behave autonomously without frequently repeating specific behavioral patterns.

Keywords: Camera image, Autonomous behavior, Machine learning

DOI: 10.54941/ahfe1005638

Cite this paper:

​