Examining the interface of neuroscience and robotics, developing methods for control and tactile feedback in prosthetic hands

我的研究结合了两个不同领域的发展。在计算神经科学和计算机科学中，尖峰神经网络(SNN)被称为第三代神经网络，因为它在前馈结构中利用了时间域。与以前的网络结构相比，它们显著减少了所需的计算资源，为探索我们的世界提供了一种新的技术。在触觉机器人领域，布里斯托尔的研究团队率先开发了一种名为TacTip的3D打印手指状结构。硅胶模具与表面接触时会凹陷，目前的深度学习技术在实验室中用于检测表面纹理。为了探索通过假肢进行人类机器人交互的可行性，我的研究目标是将SNN的低计算能力与假手上的TacTip的缩小版本相结合。首先，我正在开发一个可推广的无监督SNN，并研究该网络在检测独特的手指运动(如敲击、边缘检测和滑动)方面的成功。在我的博士生涯中，对未来工作的一些想法包括为用户提供触觉反馈，以及跨多指触控提示进行整合。

My research combines developments in two difference fields. In computational neuroscience and computer science, Spiking Neural Networks (SNN) are heralded as the third generation of neural networks due to their utilization of the temporal domain in a feedforward structure. Coupled with a significant reduction in required computational resources relative to previous network structures, they offer a new technique to explore our world. In the field of tactile robotics, the research team here at Bristol have pioneered a 3D printed finger-like structure known as the TacTip. The silicone mould depresses upon contact with a surface and current deep learning techniques are utilised in the lab to detect surface texture.In a pursuit to explore the feasibility of human robot interaction via prosthesis, my research aims to incorporate the low-computational power of SNN with a scaled-down version of the TacTip on a prosthetic hand. To begin, I am developing a generalizable unsupervised SNN and investigating the success of the network in detecting a unique finger motions such as tapping, edge-detection and slip. Some ideas for future work throughout my PhD progression involve haptic feedback for the user and incorporation across multi-finger TacTips.