A bioinspired analogous nerve towards artificial intelligence

The development of artificial sensory neural systems is crucial for restoring touch perception in disabled individuals, creating hybrid bioelectric reflex arcs, and building interactive robotic feedback. While advancements have been made in bionic sensors and signal transmission methods, most existing devices use discrete elements, leading to complex interconnections, circuitry, and signal transmission issues. Additionally, power consumption remains a challenge. This research aims to address these limitations by proposing an 'all-in-one' bionic sensory transmission nerve, the APT nerve, that integrates mechanosensation and signal transmission. The APT nerve is designed to mimic the functions of biological sensory neurons, converting mechanical stimulation into electrical signals and recognizing the location of the stimulation without multiple sensing units. Its design is inspired by the structure and function of biological sensory neurons and synapses, aiming for flexibility, low power consumption, and high performance. The use of a fiber-based paper substrate contributes to the device's flexibility, while the separate electrical double-layer structure ensures stability and energy efficiency. The goal is to create a device capable of sophisticated interactions for applications in advanced prosthetics and robotics.

Existing research has made progress in creating individual components of artificial sensory systems. Bionic sensors, such as those using giant magneto-impedance materials, offer high sensitivity. Ionic cables have been proposed for high-speed, long-distance signal transmission. Artificial spiking transistors have demonstrated spatiotemporal signal processing. However, these advancements are often implemented with discrete elements, resulting in complex systems with potential signal transmission issues and high energy consumption. The authors highlight the need for an integrated approach to overcome these limitations.

The APT nerve is fabricated using a simple process. Two pieces of adhesive tape create a hollow pattern on graph paper. An 8B pencil is used to draw a conductive graphite film within the pattern. This creates two conductive films, which are then taped together face-to-face with a spacer to form the APT nerve. The spacer controls the sensitivity and response range. The device's performance is characterized by testing its response to mechanical stimulation with varying spacer thickness, active layer width, and active layer length. The response resistance is measured and analyzed to determine the location of the stimulation. The APT nerve's robustness and durability are evaluated through extensive testing (>10,000 cycles). A driver circuit is designed to interface the APT nerve with a computer, enabling multifunctional touch interactions. The circuit includes a linear voltage transformer, an analog-to-digital converter (ADC), and a microcontroller (Arduino Leonardo) that processes the signals and sends commands to a computer for specific actions. The APT nerve is tested in various configurations (linear, L-shaped, square) for different applications. Machine learning aspects are explored, analyzing characteristic parameters like touching location, holding time, and latency interval to classify different mechanical stimulations.

The APT nerve exhibits several key features: (1) **Integrated functionality:** It combines mechanosensation, signal transmission, and location recognition in a single device. (2) **Low power consumption:** It consumes energy only during mechanical stimulation. (3) **High performance:** It has a rapid response time (<21 ms), high robustness, and excellent durability (>10,000 tests). (4) **Regional differentiation:** The response signal is highly dependent on the location of the mechanical stimulation, enabling spatiotemporal dynamic logic similar to biological neural networks. (5) **Multifunctional capabilities:** It can be used for playing music, controlling two-dimensional positioning, and handling rotation. (6) **Flexible and adaptable design:** It can be fabricated in various shapes and sizes (linear, L-shaped, square) to accommodate different applications. (7) **Robustness and durability:** The device remains functional even after partial damage. (8) **Simplified signal processing:** The non-pixelated sensing characteristic simplifies signal processing. (9) **Spatiotemporal dynamic logic:** Enables time-oriented security applications. The linear relationship between response resistance and stimulation location is demonstrated experimentally, showcasing the device's ability to pinpoint the touch location accurately. The experiments demonstrate the successful application of the APT nerve in playing music, 2D positioning control, and handling rotation, showcasing its versatility and potential for advanced applications.

The APT nerve's integrated design addresses significant challenges in creating artificial sensory neural systems. The all-in-one structure simplifies the system, reducing complexity and improving efficiency compared to systems with discrete components. The low power consumption, robustness, and durability are key advantages for practical applications. The ability to mimic the spatiotemporal dynamic logic of biological neural networks opens possibilities for sophisticated learning and control algorithms. The demonstrated applications (music, 2D control, rotation) highlight the potential of the APT nerve for use in neuroprosthetics and robotics. The simplicity of the fabrication method makes the APT nerve scalable and cost-effective, paving the way for wider adoption.

This research successfully demonstrates a novel, all-in-one bionic artificial nerve, the APT nerve. Its integrated design, high performance, and adaptability make it a promising candidate for advanced applications in neuroprosthetics and robotics. Future research could explore the use of different functional materials to enhance the device's properties (e.g., stretchability, transparency), further refine the machine learning algorithms for advanced control, and explore its integration with biological systems for improved human-machine interfaces.

The current design uses a paper-based substrate, which might limit its long-term stability and biocompatibility in certain environments. While the device shows excellent durability in testing, further long-term studies are necessary to validate its performance in real-world conditions. The current applications are proof-of-concept demonstrations; more research is required to integrate the APT nerve seamlessly into complex robotic systems and neuroprosthetics.