As the basis of Tactile Internet, remote haptic display has been made possible with the development of ultra-reliable low-latency communication in 5G. In this study, edge learning is employed to enable realistic haptic display and stable remote communication. We propose a double-loop control algorithm, which merges decoupling and PID neural network, for magnetic field generation of the electromagnetic haptic device. In addition, a supervised bidirectional LSTM network is constructed for online haptic prediction during remote interaction, thus complementing the missing haptic data on account of time delay and packet loss in network communications. Experiments have been conducted on the built remote haptic display system, where data streams from sensors are gathered, stored, and forwarded in real time. The results show that dynamic and accurate haptic display is achieved through our magnetic field control algorithm for the haptic device, and the error of haptic prediction by step is less than 0.01N. The conclusion is that the service sustainability of remote haptic display can be guaranteed by edge learning effectively.

中文翻译：

---

支持边缘学习的真实触摸和稳定通信，用于远程触觉显示


作为触觉互联网的基础，随着 5G 超可靠低延迟通信的发展，远程触觉显示已成为可能。在这项研究中，采用边缘学习来实现逼真的触觉显示和稳定的远程通信。我们提出了一种融合解耦和 PID 神经网络的双环控制算法，用于电磁触觉设备的磁场生成。此外，构建了一个有监督的双向 LSTM 网络，用于远程交互过程中的在线触觉预测，从而补充了由于网络通信中的时间延迟和丢包而丢失的触觉数据。实验已经在构建的远程触觉显示系统上进行，其中来自传感器的数据流被实时收集、存储和转发。结果表明，通过我们的触觉设备磁场控制算法，实现了动态、准确的触觉显示，并且触觉预测的分步误差小于0.01N。结论是边缘学习可以有效保证远程触觉显示的服务可持续性。