面向机械臂的抓取位姿检测

Grasping Pose Detection for Robotic Arms

针对结构化环境和单一已知目标的抓取方面,搭载末端夹持器机械臂抓取技术取得了显著进展。但当工作场景环境不明、工作目标形状各异时,机械臂容易受环境光照变化、物体遮挡的影响,为此,本文深入研究了机械臂抓取系统流程和抓取方法,设计了基于卷积神经网络自主检测机械臂抓取位姿的方法。

Significant progress has been made in the grasping technology of robotic arms equipped with end clamps for structured environments and single known targets.However,when the working environment is unknown and the shapes of work targets are different,robotic arm is easily affected by changes in environmental lighting and object occlusion.Therefore,this paper in-depth studies the process and grasping methods of the robotic arm grasping system,and designs a method based on convolutional neural networks to autonomously detect the grasping posture of the robotic arm.