Triplet Network based Few Shot Outlier DetectionSystem in Robot Task Planning for reduced failures/unwanted Executions

Authors

  • MidhunM. S
  • James Kurian

DOI:

https://doi.org/10.46243/jst.2022.v7.i04.pp136-142

Keywords:

Triplet Networ, Task planning, Robotics, Outlier detection, Few-shot leaning, One-shotlearning

Abstract

Human/systems managing the robot may make errors, resulting in a significant loss.A robotic task outlier identification approach for serial manipulator setups to avoid such outliers. The suggested work generates robot tasks in the first stage by recording the joint values utilised as the proposed dataset. Then, the metric learning-based Triplet model with convolutional feature learning layers is presented for few-shot feature learning in robot tasks. Each job is represented by an n-dimensional vector, where n represents the robot's degrees of freedom. The robot work comprises about 1500 characters from the Omniglot dataset; therefore, drawing characters from several languages on a canvas is chosen as a test

Downloads

Published

2022-06-30

How to Cite

MidhunM. S, & Kurian, J. (2022). Triplet Network based Few Shot Outlier DetectionSystem in Robot Task Planning for reduced failures/unwanted Executions. Journal of Science & Technology (JST), 7(4), 136–142. https://doi.org/10.46243/jst.2022.v7.i04.pp136-142