Robot Autonomy through Learning from Multi-Camera Images and Human Selection Behavior

Abstract

The COVID-19 pandemic, which began in early 2020, highlighted the need for technologies that can mitigate the risks of human exposure during infectious disease outbreaks. Given the ongoing threat of emerging pandemics, it is crucial to develop robotic systems that can be remotely operated by humans and eventually achieve autonomous behavior through learning from such interactions. As a fundamental study in this direction, this paper presents a method for enabling robots to autonomously operate in environments. The proposed system integrates real-time and past images from multiple cameras and learns human selection behavior based on these images to enable autonomous decision-making. Experimental results demonstrate that the proposed system achieves significantly longer autonomous operation without collisions compared to the author’s previous approach.

Keywords: Multi camera images, Biped robot, Autonomous behavior, Machine learning

DOI: 10.54941/ahfe1006250