基于多模态感知的机械臂遥操作系统设计与实现

Design and Implementation of a Robotic Arm Teleoperation System Based on Multimodal Perception

摘要: 针对传统手部动作感知系统依赖单一传感模态、穿戴舒适性差等问题，设计了一种轻量化、无线化的多模态感知系统，并将其应用于机械臂遥操作。该系统由集成弯曲与压力传感器的柔性数据手套及八通道前臂肌电手环构成，通过双蓝牙架构实现运动学与生理学数据的亚毫秒级同步采集。在遥操作控制中，利用手环内置惯性测量单元(IMU)解算前臂姿态，控制机械臂空间位姿；基于数据手套手势识别驱动灵巧手完成抓取动作。实验结果表明，该系统在抓取任务中成功率达100%，验证了其在精细操作与实时交互中的可靠性与有效性。为自然人机交互提供了一种轻量化、高集成度的技术方案。

Abstract: To address the limitations of single-modal hand motion perception systems, this paper designs a lightweight, wireless multimodal sensing system and applies it to robotic arm teleoperation. The system consists of a flexible data glove integrated with bend and pressure sensors, and an eight-channel forearm EMG wristband. A dual-Bluetooth architecture enables sub-millisecond synchronous acquisition of kinematic and physiological data. In the teleoperation framework, the wristband’s IMU estimates forearm posture to control the robotic arm’s spatial position, while gesture recognition from the data glove drives the dexterous hand for grasping. Experiments show a 100% success rate in grasping tasks and a response latency of about 120 ms in dynamic interactions, validating the system’s reliability and real-time performance. This work offers a lightweight, highly integrated solution for natural human-robot interaction.