融合卷积神经网络和Transformer的轻量化抓取检测算法OA
Lightweight Grasp Detection Algorithm via CNN-Transformer Fusion
面向非结构化与杂乱场景下的实时抓取任务,提出位置感知的轻量化抓取检测网络(PAG-Net).算法在 MBConv(mobile inverted bottleneck block convolution)中嵌入 PEG(positional encoding generator)以显式注入空间位置信息,并通过增强特征引入改进的 LowFormer(low-resolution transformer)进一步提升网络的计算效率.网络以像素级生成方式同时预测抓取质量、角度与夹爪宽度,有效提升抓取检测的精度与速度.实 验 表 明,PAG-Net 在 Cornell 数据集上的准确率达到 98.8%,在Jacquard 数据集上达到 96.1%.在 PyBullet仿真测试中,PAG-Net在杂乱环境中实现了 94%的抓取成功率,证明了该网络在复杂场景中的鲁棒性.
To address real-time grasping tasks in unstructured and cluttered environments,we propose a position-aware lightweight grasp detection network,PAG-Net.The algorithm embeds PEG(positional encoding generator)into MBConv(mobile inverted bottleneck block convolution)to explicitly inject spatial position information,and further improves the network's computational efficiency by introducing an improved LowFormer(a low-resolution transformer)for feature enhancement.The network simultaneously predicts grasp quality,angle,and gripper width in a pixel-wise generation manner,effectively improving the accuracy and speed of grasp detection.Experimental results show that PAG-Net achieves an accuracy of 98.8%on the Cornell dataset and 96.1%on the Jacquard dataset.In PyBullet simulation tests,PAG-Net achieves a 94%grasping success rate in cluttered environments,proving the robustness of the proposed network in complex scenes.
胡平;李浩;张靖
安徽工程大学计算机与信息学院,安徽 芜湖 241000安徽工程大学计算机与信息学院,安徽 芜湖 241000长三角哈特机器人产业技术研究院,安徽 芜湖 241000
信息技术与安全科学
机器人抓取深度学习机械臂
robot graspingdeep learningrobotic arm
《信息与控制》 2026 (4)
714-726,13
国家自然科学基金项目(62302011)安徽省重点研究与开发计划项目(20250718KTRD0020)安徽省教育厅高校科学研究重点项目(2024AH050132)
评论