查找使用阿拉丁产品的科研文献

Dynamically Reversible Filament Networks Enabling Programmable In-Sensor Memory for High-Precision Neuromorphic Interactions

Lei Liu,Shifan Yu,Yijing Xu,Hongyu Chen,Huasen Wang,Wansheng Lin,Yu Hu,Zijian Huang,Chao Wei,Yuchen Lin,Ziquan Guo,Tingzhu Wu,Jianghui Zheng,Zhong Chen,Yuanjin Zheng,Xinqin Liao

Embodied intelligent tactile systems represent a groundbreaking paradigm for autonomous agents, facilitating dynamic perception and adaptation in unstructured environments. Traditional von Neumann architectures suffer from inefficiencies due to the separation of sensing and memory units, where mechanical relaxation is often overlooked as non-informative noise rather than utilized as a computational resource. The transition dynamics from mechanical stimulation to memory encoding and their potential in neuromorphic interactions remain largely unexplored. Here, we present a transformative breakthrough in the seamless integration of sensing and memory (SMI) within a single device through programmable tactile memory. Utilizing polyborosiloxane (PBS) filament networks with dynamically reversible boron-oxygen and hydrogen bonds, the design enhances adhesion and energy dissipation. It enables pressure-induced electrically readable memory states with tunable retention times (260 ms to 63.9 s) and 99.6% linearity, supporting applications, such as threshold triggering, biomimetic pain perception, and motion recognition. The SMI sensor's in-sensor memory and logic functions facilitate intelligent control, while its memory retention capabilities enable pain visualization and action-driven modulation. Additionally, the spatiotemporal tactile memory achieves high-precision motion recognition (98.33%) without relying on continuous time-series data. This work introduces a novel mechanism for constructing SMI devices, advancing the development of intelligent neuromorphic tactile systems.