Scholarlink

The paper outlines the integration of a motion-tracking system and robotic arm to intercept and catch an object thrown by a human. The objective is to achieve a precise position for the robot before the ball enters it’s workspace. Predicting the trajectory of the thrown object is an essential task. Information on the object’s motion during the early stage of flight is obtained from the motion-tracking system. Trajectory estimation uses two methods: an analytical equation for an oblique projectile and a trained deep neural network for time-series data prediction. The software implementation of this solution has been tested in real-world conditions, and results show that the proposed pipeline is able to catch the ball in flight.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Robot Learning to Catch Objects in Flight Using an LSTM Deep Neural Network

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Lazar Milić,
  • Jefimija Borojević,
  • Tanja Berisavljević,
  • Srđan Savić,
  • Mirko Raković

摘要

The paper outlines the integration of a motion-tracking system and robotic arm to intercept and catch an object thrown by a human. The objective is to achieve a precise position for the robot before the ball enters it’s workspace. Predicting the trajectory of the thrown object is an essential task. Information on the object’s motion during the early stage of flight is obtained from the motion-tracking system. Trajectory estimation uses two methods: an analytical equation for an oblique projectile and a trained deep neural network for time-series data prediction. The software implementation of this solution has been tested in real-world conditions, and results show that the proposed pipeline is able to catch the ball in flight.