Open AccessDevelopment and application of substation intelligent inspection robot supporting deep learning accelerating
Author(s) -
Lei Su,
Xiaoping Yang,
Boyuan Cao,
Yun Wang,
Xinjia Li,
Wenlian Lu
Publication year - 2021
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1754/1/012170
Subject(s) - deep learning , artificial intelligence , computer science , robot , inference , object detection , segmentation , fault (geology) , task (project management) , fault detection and isolation , embedded system , real time computing , engineering , systems engineering , seismology , actuator , geology
Substation inspection robots usually have low computation capability to run deep learning models. In this paper, a substation intelligent inspection robot that can support real time deep learning models is developed. The Nvidia Jetson TX2 module is the accelerating hardware module of the robot, and TensorRT is the software framework for deep learning model inference accelerating. The robot can satisfy the computational requirement of deep learning based applications, and operate in very low energy consumption. Test results on a fault diagnostic task based on object detection and instance segmentation show that with deep learning accelerating, not only high detection accuracy is achieved, but also inference time is short enough for real time fault diagnostic application.