Navegação social com restrições baseada em uma nova percepção robocêntrica

Abstract

The advancement of technologies capable of operating autonomously with precision levels close to the desired is on the rise. The application of machine learning techniques introduces a new dimension to computing and robotics, both directly benefiting from the presence of intelligent machines. In this project, we promoted an adaptation in object detection to enhance social navigation in autonomous robotic systems. To achieve this, we categorized identified objects based on their physical characteristics (living dynamic, inanimate dynamic, living static, and inanimate static ) in order to classify them in a social context. Consequently, a safety area is generated around these objects, which must be respected during their interactions. The results of validation methods were presented. Thus, it was concluded that the presented framework is feasible to be implemented with a focus on social navigation in mobile robots.