Scholarlink

Deep learning (DL) techniques leverage complex neural networks to solve intricate data analysis problems, exhibiting comparative proficiency to that of domain experts. DL stands as a remarkable source of innovation and advancement in the realm of artificial intelligence (AI). Additionally, reinforcement learning (RL) endeavors to identify the most optimal behavioral patterns of a system by exploring and acquiring knowledge through successive interactions with the surrounding environment. The DL and RL enable the effective resolution of intricate problems in system modeling, particularly in solving inverse kinematics of robot manipulation. This paper presents a deep reinforcement learning (DRL) approach specifically designed for the inverse kinematics of a 5-axis Mitsubishi RV-2AJ robotic arm. The developed approach employs advanced techniques from the field of DRL to effectively optimize the arm's movements and achieve an accurate inverse kinematics solution for the robot. The proposed DRL approach for inverse kinematics was evaluated against alternative classical control methodologies such as the Denavit–Hartenberg (DH) method. The obtained results exhibited the superior efficiency and effectiveness of the DRL approach.

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

Study of Inverse Kinematics Solution for a 5-Axis Mitsubishi RV-2AJ Robotic Arm Using Deep Reinforcement Learning

  • 章节

  • 发表日期：

    2025年

查看全文
  • Zied Ben Hazem,
  • Nivine Guler,
  • Walid El Fezzani

摘要

Deep learning (DL) techniques leverage complex neural networks to solve intricate data analysis problems, exhibiting comparative proficiency to that of domain experts. DL stands as a remarkable source of innovation and advancement in the realm of artificial intelligence (AI). Additionally, reinforcement learning (RL) endeavors to identify the most optimal behavioral patterns of a system by exploring and acquiring knowledge through successive interactions with the surrounding environment. The DL and RL enable the effective resolution of intricate problems in system modeling, particularly in solving inverse kinematics of robot manipulation. This paper presents a deep reinforcement learning (DRL) approach specifically designed for the inverse kinematics of a 5-axis Mitsubishi RV-2AJ robotic arm. The developed approach employs advanced techniques from the field of DRL to effectively optimize the arm's movements and achieve an accurate inverse kinematics solution for the robot. The proposed DRL approach for inverse kinematics was evaluated against alternative classical control methodologies such as the Denavit–Hartenberg (DH) method. The obtained results exhibited the superior efficiency and effectiveness of the DRL approach.