Artificial sensory neurons: Bridging biology and technology for next-generation intelligent systems

The rise of artificial intelligence (AI) has transformed industries like healthcare, transportation, and manufacturing. However, the traditional von Neumann architecture limits computational speed and energy efficiency due to the separation of computing and storage units, making it less suitable for energy-sensitive applications like edge computing and the Internet of Things (IoT). Brain-inspired computing, modeled on biological neural networks, offers a solution through efficient parallel processing, addressing these limitations.

Spiking neural networks (SNNs), known for energy efficiency and spatiotemporal processing, communicate via discrete spikes generated when signals exceed a threshold. Efficient conversion of environmental analog signals into spike signals is critical for their application. Traditional methods like analog-to-digital converters (ADCs) suffer from high power consumption and complexity, limiting their use in edge scenarios.

Zhong Shuai from the Guangdong Institute of Intelligent Science and Technology and Zhang Yishu from Zhejiang University have made groundbreaking contributions to the field of artificial sensory neurons (ASNs). They first elucidated the transduction mechanisms of various biological signals—such as vision, touch, thermal perception, taste, smell, and hearing—drawing on a deep understanding of neuroscience principles. By bridging biological insights with advanced device engineering, they provided a systematic analysis of devices used to construct ASNs, including memristors, single-transistor latch-based field-effect transistors (STLFETs), and 2D memtransistors.

Artificial sensory neurons (ASNs) have emerged as an alternative, leveraging devices like memristors, phase-change memory (PCM), and ferroelectric field-effect transistors (FeFETs) to efficiently convert continuous signals into spikes. These devices offer scalability, low power consumption, and fast responses, surpassing the limitations of conventional CMOS technology. Their research highlights the critical role of ASNs in translating environmental signals into electrical spikes, a cornerstone for brain-inspired computational systems. Zhong and Zhang also categorized advancements across diverse ASNs, including artificial tactile, thermal, auditory, gustatory, olfactory, visual, biochemical, and multimodal sensory neurons, and explored their applications in areas such as machine learning, nociception, collision avoidance, and neural interfaces. By proposing key performance metrics and assessing application feasibility, they have provided a comprehensive roadmap for developing high-performance ASNs, significantly advancing the field of brain-inspired technologies and paving the way for innovative sensory systems.

Energy consumption, a key consideration, can be reduced by integrating sensing and encoding functionalities, optimizing materials, and enhancing device design. Balancing sensitivity and dynamic range is crucial for practical applications, and innovative materials like ferroelectric and porous composites show promise in achieving this balance. Linearity, response time, and resolution are also critical. Linearity ensures proportionality between input and output, while response time affects real-time performance. Resolution determines the smallest detectable input change, essential for precision applications. These parameters can be improved through material and structural innovations, although challenges remain. Stability, particularly under environmental variations, is vital for long-term reliability. For biodegradable electronics, adaptability to specific environments is essential, with thermal stability and application-specific lifetimes being key research areas.

ASNs offer advantages over traditional ADCs in cost and energy efficiency, with potential applications in AI, bionic prosthetics, and humanoid robotics. However, challenges such as poor thermal stability, high power consumption, and fabrication complexity need addressing through advancements in material science and manufacturing techniques. Future ASNs should enhance compactness, response time, and dynamic range to better mimic human sensory systems.

Multimodal sensory neurons capable of detecting diverse stimuli represent a key frontier. Small-scale ASN arrays have demonstrated success in tactile and visual encoding, but issues like signal interference and thermal effects must be resolved to enable higher density and resolution. For practical use, ASNs must integrate with neural networks and actuators, requiring optimized interface circuits and efficient algorithms for spiking neural networks. These systems hold promise for real-time, resource-constrained environments, enabling multimodal fusion and energy-efficient processing.

ASN research, though nascent, stands to benefit from interdisciplinary collaboration, accelerating advancements in neuromorphic hardware and offering innovative solutions for intelligent autonomous systems. With further development, ASNs could revolutionize brain-inspired computing, providing the foundation for next-generation technologies.