The National Natural Science Foundation of China (General Program, Key Program, Major Research Plan)
触觉智能感知是当前研究的热点问题之一。然而,大规模触觉数据集的缺乏限制了机器人触觉感知领域的发展,解决问题的关键在于构建覆盖手掌的高时空分辨率触觉压力传感器系统。本文构建了脑启发的触觉传感系统(Brain-inspired tactile sensing system, BITSS),以高时空分辨率对触觉压力信息进行获取,并实现了基于脉冲事件的触觉感知。受皮肤触觉感受器启发,BITSS使用神经形态模型对触感压力信号进行脉冲编码,实现了两种触觉感受器神经元的模拟。这项工作的创新之处在于实现了对触觉压力信息的脉冲编码,并基于脉冲信号对抓握过程和抓握对象进行触觉感知。实验结果表明,BITSS模拟的神经电活动可以解码出抓握状态的低维空间。在10种日常物体的分类任务中,我们提供了基于脉冲的贝叶斯分类器,分类精度达到94%,并具有较快的执行速度。以上结果验证了BITSS对触感压力信号的时空编码能力。BITSS利用了生物神经网络脉冲编码的时空编码特性,为基于脉冲神经网络的力反馈机器人和触感神经修复技术提供了支持。
Tactile intelligent perception is one of the hot issues in current researches. However, The lack of large-scale tactile data sets limits the development of the field of tactile perception. The key to solving the problem is to build a high-temporal-resolution tactile pressure sensor system that covers the palm. This paper constructs a brain-inspired tactile sensing system (BITSS), which acquires tactile pressure information with high spatiotemporal resolution and realizes tactile perception based on spike events. Inspired by the skin"s tactile receptors, BITSS uses a neuromorphic model to spike-encode tactile pressure signals and realizes the simulation of two types of tactile receptor neurons. The innovation of this work lies in the realization of pulse encoding of tactile pressure information and the tactile perception of the grasping process and the grasped object based on the pulse signal. The experimental results show that the neuroelectric activity simulated by BITSS can decode the low-dimensional space of the grasping state. In the classification tasks of 10 daily objects, we provide aa spike-based Bayesian classifier with a classification accuracy of 94% and a fast execution speed. The above results verify the spatiotemporal coding ability of BITSS for tactile pressure signals. BITSS utilizes the spatio-temporal coding characteristics of biological neural network pulse coding to support force feedback robots and tactile nerve repair technology based on pulse neural networks.