Within the IoT 2.0 and the Industry 4.0 paradigms, the transition from cloud to edge computing is vital for homogeneous and universal data access throughout smart connected devices and products lifecycles. This calls for more advanced edge devices, which often struggle with power constraints; one of the key players in this challenge is the Neuromorphic technology, inspired by the most advanced and power-efficient sensor data analytics system — the human brain.
Although initially intended for brain simulations, the adoption of the emerging Neuromorphic technology is more and more appealing in fields such as IoT edge devices, Industry 4.0, Biomedical, HPC, and Robotics. This trend is confirmed by the effort of several companies in developing Neuromorphic architectures and software tools that are opening the way to a new family of hybrid Neuromorphic/Digital IoT devices.
Neuromorphic computational paradigms and hardware architectures are now mature enough to play an important role in IoT applications running on-edge, because of their ability to learn and adapt to ever-changing conditions and tasks while respecting limited power requirements.
Several state-of-the-art benchmark applications have proved that Neuromorphic solutions, since brain-inspired, provide better scalability than traditional multi-core architectures, and are especially suited for low-power and adaptive applications required to analyse data in real-time.
In the literature emerged several Neuromorphic tools:
• Analog/Digital HW architectures (Loihi, SpiNNaker, DYNAP-SEL);
• SNN simulators (Nest, GeNN, Nengo);
• SNN computing primitives (multi-compartment models, SNAN-astrocytic, Stick);
• Learning methods (LMU, BPTT, L2L, e-prop);
• Optimization tools (EONS);
• Encoding techniques (Rate-coding, Temporal-coding);
• APIs and compilers (PyNN, SpyTorch, snnTorch);
• Dedicated EDA tools (FPAA, pyNAVIS, SNN-TB);
• Neuromorphic sensors (DVS, Silicon-Cochlea, Tactile);
• Benchmarks (SNN-Classifier, Sudoku, Robotic control);
• Integrated frameworks (NEF, Intel-Lava, Fugu, NeuCube).
However, the weak standardisation of Neuromorphic components, tools, and frameworks makes it challenging to define the engineering process for developing and orchestrating hybridized Neuromorphic/Digital System of Systems deployable in real-world application scenarios.
The Research Topic will focus on various theoretical and practical aspects of different Neuromorphic setups for facilitating the adoption of Neuromorphic technology into the design of System of Systems products and algorithms for IoT applications.
Relevant topics include (but are not limited to):
• Coupling of conventional and Neuromorphic computing systems for deploying reliable heterogeneous edge solutions and SNN-oriented tinyML applications;
• Theory and applications of Neuromorphic encoding, signalling, and decoding of Spatio-temporal data streams collected with standard sensors;
• Neuromorphic audio/video/tactile/motion/environmental sensors, real-time processing, sensory fusion, and on-line classification;
• Neuromorphic HW/SW architectures for low-power computation and cognition;
• Biologically-inspired computational primitives, learning, and adaptation rules for specific problems;
• Advanced and user-friendly SDK, API, compilers, and frameworks for design, optimize, and simulate Neuromorphic solutions;
• Neuromorphic solutions for replacing standard data analysis techniques (PID control, signal filtering, spectral analysis, pattern matching, CSP/COP);
• Integrated frameworks for multi-level design and benchmarking of Neuromorphic HW/SW computing systems.
The main idea is to collect original research — articles, reviews, and commentaries — about Neuromorphic tools, toolchains, programming frameworks, computational paradigms, algorithms, HW platforms, and applications, which will act as a reference for application developers involved in the design of hybrid digital/neuromorphic systems for the IoT, Biomedical, and Industrial domains.
Within the IoT 2.0 and the Industry 4.0 paradigms, the transition from cloud to edge computing is vital for homogeneous and universal data access throughout smart connected devices and products lifecycles. This calls for more advanced edge devices, which often struggle with power constraints; one of the key players in this challenge is the Neuromorphic technology, inspired by the most advanced and power-efficient sensor data analytics system — the human brain.
Although initially intended for brain simulations, the adoption of the emerging Neuromorphic technology is more and more appealing in fields such as IoT edge devices, Industry 4.0, Biomedical, HPC, and Robotics. This trend is confirmed by the effort of several companies in developing Neuromorphic architectures and software tools that are opening the way to a new family of hybrid Neuromorphic/Digital IoT devices.
Neuromorphic computational paradigms and hardware architectures are now mature enough to play an important role in IoT applications running on-edge, because of their ability to learn and adapt to ever-changing conditions and tasks while respecting limited power requirements.
Several state-of-the-art benchmark applications have proved that Neuromorphic solutions, since brain-inspired, provide better scalability than traditional multi-core architectures, and are especially suited for low-power and adaptive applications required to analyse data in real-time.
In the literature emerged several Neuromorphic tools:
• Analog/Digital HW architectures (Loihi, SpiNNaker, DYNAP-SEL);
• SNN simulators (Nest, GeNN, Nengo);
• SNN computing primitives (multi-compartment models, SNAN-astrocytic, Stick);
• Learning methods (LMU, BPTT, L2L, e-prop);
• Optimization tools (EONS);
• Encoding techniques (Rate-coding, Temporal-coding);
• APIs and compilers (PyNN, SpyTorch, snnTorch);
• Dedicated EDA tools (FPAA, pyNAVIS, SNN-TB);
• Neuromorphic sensors (DVS, Silicon-Cochlea, Tactile);
• Benchmarks (SNN-Classifier, Sudoku, Robotic control);
• Integrated frameworks (NEF, Intel-Lava, Fugu, NeuCube).
However, the weak standardisation of Neuromorphic components, tools, and frameworks makes it challenging to define the engineering process for developing and orchestrating hybridized Neuromorphic/Digital System of Systems deployable in real-world application scenarios.
The Research Topic will focus on various theoretical and practical aspects of different Neuromorphic setups for facilitating the adoption of Neuromorphic technology into the design of System of Systems products and algorithms for IoT applications.
Relevant topics include (but are not limited to):
• Coupling of conventional and Neuromorphic computing systems for deploying reliable heterogeneous edge solutions and SNN-oriented tinyML applications;
• Theory and applications of Neuromorphic encoding, signalling, and decoding of Spatio-temporal data streams collected with standard sensors;
• Neuromorphic audio/video/tactile/motion/environmental sensors, real-time processing, sensory fusion, and on-line classification;
• Neuromorphic HW/SW architectures for low-power computation and cognition;
• Biologically-inspired computational primitives, learning, and adaptation rules for specific problems;
• Advanced and user-friendly SDK, API, compilers, and frameworks for design, optimize, and simulate Neuromorphic solutions;
• Neuromorphic solutions for replacing standard data analysis techniques (PID control, signal filtering, spectral analysis, pattern matching, CSP/COP);
• Integrated frameworks for multi-level design and benchmarking of Neuromorphic HW/SW computing systems.
The main idea is to collect original research — articles, reviews, and commentaries — about Neuromorphic tools, toolchains, programming frameworks, computational paradigms, algorithms, HW platforms, and applications, which will act as a reference for application developers involved in the design of hybrid digital/neuromorphic systems for the IoT, Biomedical, and Industrial domains.
