RRAM-based CAM combined with time-domain circuits for hyperdimensional computing

Content addressable memory (CAM) for search and match operations demands high speed and low power for near real-time decision-making across many critical domains. Resistive RAM (RRAM)-based in-memory computing has high potential in realizing an efficient static CAM for artificial intelligence tasks, especially on resource-constrained platforms. This paper presents an XNOR-based RRAM-CAM with a time-domain analog adder for efficient winning class computation. The CAM compares two operands, one voltage and the second one resistance, and outputs a voltage proportional to the similarity between the input query and the pre-stored patterns. Processing the summation of the output similarity voltages in the time-domain helps avoid voltage saturation, variation, and noise dominating the analog voltage-based computing. After that, to determine the winning class among the multiple classes, a digital realization is utilized to consider the class with the longest pulse width as the winning class. As a demonstrator, hyperdimensional computing for efficient MNIST classification is considered. The proposed design uses 65 nm CMOS foundry technology and realistic data for RRAM with total area of 0.0077 mm2, consumes 13.6 pJ of energy per 1 k query within 10 ns clock cycle. It shows a reduction of ~ 31 × in area and ~ 3 × in energy consumption compared to fully digital ASIC implementation using 65 nm foundry technology. The proposed design exhibits a remarkable reduction in area and energy compared to two of the state-of-the-art RRAM designs.

RRAM-Based CAM Combined With Time-Domain Circuits for Hyperdimensional Computing

Content addressable memory (CAM) for search and match operations demands high speed and low power for near real-time decision-making across many critical domains. Resistive RAM-based in-memory computing has high potential in realizing an efficient static CAM for artificial intelligence tasks, especially on resource-constrained platforms. This paper presents an XNOR-based RRAM-CAM with a time-domain analog adder for efficient winning class computation. The CAM compares two operands, one voltage and the second one resistance, and outputs a voltage proportional to the similarity between the input query and the pre-stored patterns. Processing the summation of the output similarity voltages in the time-domain helps avoid voltage saturation, variation, and noise dominating the analog voltage-based computing. After that, to determine the winning class among the multiple classes, a digital realization is utilized to consider the class with the longest pulse width as the winning class. As a demonstrator, hyperdimensional computing for efficient MNIST classification is considered.The proposed design uses 65nm CMOS foundry technology and realistic data for RRAM with total area of 0.0077 mm2 , consumes 13.6 pJ of energy per 1k query within 10 ns clock cycle for 10 classes. It shows a reduction of ∼ 31× in area and ∼ 3× in energy consumption compared to fully digital ASIC implementation using 65nm foundry technology. The proposed design exhibits a remarkable reduction in area and energy compared to two of the state-of-the-art RRAM designs.

Neural control of robot manipulators considering motor voltage saturation: performance evaluation and experimental validation

Purpose This paper aims to design a neural controller based on radial basis function networks (RBFN) for electrically driven robots subjected to constrained inputs. Design/methodology/approach It is assumed that the electrical motors have limitations on the applied voltages from the controller. Due to the universal approximation property of RBFN, uncertainties including un-modeled dynamics and external disturbances are represented with this powerful neural network. Then, the lumped uncertainty including the nonlinearities imposed by actuator saturation is introduced and a mathematical model suitable for model-free control is presented. Based on the closed-loop equation, a Lyapunove function is defined and the stability analysis is performed. It is assumed that the electrical motors have limitations on the applied voltages from the controller. Findings A comparison with a similar controller shows the superiority of the proposed controller in reducing the tracking error. Experimental results on a SCARA manipulator actuated by permanent magnet DC motors have been presented to guarantee its successful practical implementation. Originality/value The novelty of this paper in comparison with previous related works is improving the stability analysis by involving the actuator saturation in the design procedure. It is assumed that the electrical motors have limitations on the applied voltages from the controller. Thus, a comprehensive approach is adopted to include the saturated and unsaturated areas, while in previous related works these areas are considered separately. Moreover, a performance evaluation has been carried out to verify satisfactory performance of transient response of the controller.

Numerical Study of Sub-Gap Density of States Dependent Electrical Characteristics in Amorphous In-Ga-Zn-O Thin-Film Transistors

We demonstrate the effect of the sub-gap density of states (DOS) on electrical characteristics in amorphous indium-gallium-zinc (IGZO) thin-film transistors (TFTs). Numerical analysis based on a two-dimensional device simulator Atlas controlled the sub-gap DOS parameters such as tail acceptor-like states, tail donor-like states, Gauss acceptor-like states, and Gauss donor-like states in amorphous IGZO TFTs. We confirm accuracy by exploiting physical factors, such as oxygen vacancy, peroxide, hydrogen complex, band-to-band tunneling, and trap-assisted tunneling. Consequently, the principal electrical parameters, such as the threshold voltage, saturation mobility, sub-threshold swing, and on-off current ratio, are effectively tuned by controlling sub-gap DOS distribution in a-IGZO TFTs.

Lyapunov Sliding-Mode Observers With Application for Active Magnetic Bearing Operated With Zero-Bias Flux

This study deals with sliding-mode nonlinear observers for a flux-controlled active magnetic bearing (AMB) operated with zero-bias flux. The Lyapunov sliding-mode observer (LSMO) feedback designs are performed for the nonlinear AMB dynamics due to control voltage saturation. The nonlinear observers are designed to estimate the magnetic flux and rotor mass velocity. The observer designs are incorporated in equivalence implementation of the nonlinear state-feedback controller. The main design tools such as sliding-mode control, Lyapunov-based control are used in this framework. The proposed observers are verified by means of numerical simulations, and stability and effectiveness of the proposed observer-based feedback designs are shown.