Machine learning, mobile computing, and internet of things had brought about the most disruptive technological advances in the past decade. The blending of these technologies with ever-evolving artificial intelligence had reflected itself in smart devices such as mobile phones, watches, and glasses etc. With the wider spread of technologies among industries, such devices have gradually taken over a large market share in consumer electronics, with several billions of devices sold each year leading by industrial giants such as Apple and Samsung. However, digital electronics-based smart edge devices are facing the challenge of meeting demanding requirements in performance while being energy-efficient and small in size. In conventional digital hardware devices and systems, modules for sensing, memory, and processing units are designed separately. However, the sequential analog-digital conversion is integral to almost every AI task, so does the frequent and massive data transfer among these components, leading to further time latency and power consumption. Moore’s law puts a limitation in terms of solving this difficulty by scaling up of transistors, and therefore, fundamental changes to the edge computing paradigm are critical and urgently needed.
The objective of this Research Topic is to bridge near-sensor and in-sensor computing with brain-inspired computing to facilitate the development of a neuromorphic system that extracts, analyzes, and computes the sensory signals via an in-situ style, effectively boosting processing power and energy efficiency. Such integrated systems will find applications in low-power edge intelligent devices, where sensory data could be accessed swiftly to generate a decision without connecting to cloud. The near-sensor and in-sensor brain-inspired computing could be approached from multiple perspectives. At the sensory component level, novel visual sensors with light-adaptive and non-volatile responsivity could store information and compute within, reducing the cost for data transfer and conversion. At the signal processing level, circuits based on analogue and hybrid signals can be developed to remove or simplify ADC conversion process to keep good balance between computing precision and power consumption.
This Research Topic aims to bring together state-of-the-art interdisciplinary research in the above context including enabling technologies for brain-inspired in-sensor and near-sensor computing. Bio-inspired principles and engineering approaches for the fusion of intelligent sensing and computing are more than welcomed. Articles emphasizing emerging devices, circuits, architectures, algorithms and co-designs of these hierarchies will be welcomed, but will need to clearly link back to Brain-inspired Computing.
Topics of interests include but not limited to:
• Biomimetic architectures for modelling multi-modal perception and cognition neuromorphic systems
• Bio-inspired/Dedicated near/in-sensor computing circuits, architectures, algorithms, and systems
• Key workloads e.g., AI, database, graph processing, etc. taking advantage of near/in-sensor processing
• Near/in-sensor computing schemes for data management, on-chip routing and communication
• Advanced integration and packaging technologies facilitating near/in-sensor brain-inspired computing e.g., 2.5D/3D integration, chiplet, etc
• Benchmarking and workload analysis for near/in-sensor brain-inspired computing
• Issues and countermeasures for reliability, energy efficiency, fault tolerance, etc.
• EDA design and Simulation tools/platforms for near/in-sensor computing
• Multimodal sensory system integrated with artificial intelligence
• Emerging materials and devices for near/in-sensor brain-inspired computing
Machine learning, mobile computing, and internet of things had brought about the most disruptive technological advances in the past decade. The blending of these technologies with ever-evolving artificial intelligence had reflected itself in smart devices such as mobile phones, watches, and glasses etc. With the wider spread of technologies among industries, such devices have gradually taken over a large market share in consumer electronics, with several billions of devices sold each year leading by industrial giants such as Apple and Samsung. However, digital electronics-based smart edge devices are facing the challenge of meeting demanding requirements in performance while being energy-efficient and small in size. In conventional digital hardware devices and systems, modules for sensing, memory, and processing units are designed separately. However, the sequential analog-digital conversion is integral to almost every AI task, so does the frequent and massive data transfer among these components, leading to further time latency and power consumption. Moore’s law puts a limitation in terms of solving this difficulty by scaling up of transistors, and therefore, fundamental changes to the edge computing paradigm are critical and urgently needed.
The objective of this Research Topic is to bridge near-sensor and in-sensor computing with brain-inspired computing to facilitate the development of a neuromorphic system that extracts, analyzes, and computes the sensory signals via an in-situ style, effectively boosting processing power and energy efficiency. Such integrated systems will find applications in low-power edge intelligent devices, where sensory data could be accessed swiftly to generate a decision without connecting to cloud. The near-sensor and in-sensor brain-inspired computing could be approached from multiple perspectives. At the sensory component level, novel visual sensors with light-adaptive and non-volatile responsivity could store information and compute within, reducing the cost for data transfer and conversion. At the signal processing level, circuits based on analogue and hybrid signals can be developed to remove or simplify ADC conversion process to keep good balance between computing precision and power consumption.
This Research Topic aims to bring together state-of-the-art interdisciplinary research in the above context including enabling technologies for brain-inspired in-sensor and near-sensor computing. Bio-inspired principles and engineering approaches for the fusion of intelligent sensing and computing are more than welcomed. Articles emphasizing emerging devices, circuits, architectures, algorithms and co-designs of these hierarchies will be welcomed, but will need to clearly link back to Brain-inspired Computing.
Topics of interests include but not limited to:
• Biomimetic architectures for modelling multi-modal perception and cognition neuromorphic systems
• Bio-inspired/Dedicated near/in-sensor computing circuits, architectures, algorithms, and systems
• Key workloads e.g., AI, database, graph processing, etc. taking advantage of near/in-sensor processing
• Near/in-sensor computing schemes for data management, on-chip routing and communication
• Advanced integration and packaging technologies facilitating near/in-sensor brain-inspired computing e.g., 2.5D/3D integration, chiplet, etc
• Benchmarking and workload analysis for near/in-sensor brain-inspired computing
• Issues and countermeasures for reliability, energy efficiency, fault tolerance, etc.
• EDA design and Simulation tools/platforms for near/in-sensor computing
• Multimodal sensory system integrated with artificial intelligence
• Emerging materials and devices for near/in-sensor brain-inspired computing
