Abstract: A method for processing an image, comprising: generating, by a hybrid attention processing unit, a first matrix based on a feature associated with the image and a second matrix based on information associated with the feature; and processing, by the hybrid attention processing unit, the image with the first and second matrices based on a linear attention process.

Abstract: This application provides an antenna element, including a radiating structure, where the radiating structure includes a first conductive structure, a conductive layer, and a radiative metasurface, the first conductive structure surrounds at least a part of an edge of the conductive layer and is connected to the conductive layer, the radiative metasurface and the conductive layer are stacked, the radiative metasurface is located on a side that is of the first conductive structure and that faces away from the conductive layer, there is a first gap between the radiative metasurface and the first conductive structure, the radiative metasurface includes a plurality of conductive units, and there is a second gap between every two adjacent conductive units. This application further provides an antenna system including the antenna element and a communication device. The solutions of this application can improve cross polarization discrimination of an antenna.

Abstract: A system and method is provided for designing and optimizing hardware accelerators for neural networks. During a pre-design phase, rules are extracted from compilation patterns that describe conversion between neural network operators, coarse-grained operators, and fine-grained dataflow. A fast mapper for converting neural network models to coarse-grained operator descriptions and a dataflow mapper are generated. A coarse-grained design phase employs an architecture optimizer to generate plural provisional hardware accelerator designs The coarse-grained operator descriptions are simulated using a coarse-grained simulator to obtain performance metrics of each provisional accelerator design. A fine-grained design phase employs a dataflow mapper and fine-grained simulator to finalize provisional hardware accelerator designs.

Abstract: This invention proposes a GQA-LUT method, utilizing a genetic algorithm and LUT-based circuit to efficiently approximate non-linear operators in Transformers. It adaptively finds optimal solutions for various non-linear functions, outperforming conventional neural network methods. A novel rounding mutation (RM) algorithm enhances approximation accuracy during quantization, improving low-bit integer precision. The invention also introduces a LayerNorm folding strategy as a near-memory computing principle, reducing IO and energy overheads with a two-stage memory hierarchy. Additionally, an additive partial sum quantization method is proposed to reduce energy consumption by quantizing accumulated PSUMs in matrix multiplication, alongside a PSQ-APSQ grouping strategy and floating-point regularization.

Abstract: A control system, where a first isolation unit in a control system can receive a first level signal output by a first control unit and a second level signal output by a first controlled unit, and can output a first control signal to the first controlled unit. The first controlled unit can output the second level signal to the first isolation unit under an action of the first control signal. Based on this, even if the first control unit is faulty, the first isolation unit can continue to output the first control signal to the first controlled unit based on the second level signal output by the first controlled unit. This implements isolation of a fault of the first control unit.

Abstract: A control system, a sensor, a controller, and a terminal device are provided. The control system includes a first controller and a second controller. Both the first controller and the second controller are connected to the sensor. The sensor includes a first sensing module and a second sensing module. The second controller includes a sampling circuit. The sampling circuit includes a first channel and a second channel. The first channel is connected to the first sensing module, and the second channel is connected to the second sensing module. The second controller is configured to: obtain, through the first channel, a first signal detected by the first sensing module, obtain, through the second channel, a second signal detected by the second sensing module, and determine second opening information of a pedal based on the first signal and the second signal.

Abstract: The method includes: determining required virtual inertia based on a frequency change rate monitored at a grid connection point of a photovoltaic grid-connected system; obtaining values of indexes that affect virtual inertia provided by each virtual synchronous generator (VSG) control unit in a photovoltaic cluster of the photovoltaic grid-connected system; determining, by using a predetermined function equation, a function value corresponding to each index; and determining an allocation proportion of the required virtual inertia in each VSG control unit based on the function value corresponding to each index. A problem of difficult conversion between a plurality of kinds of performance during operation of a power system is resolved, the frequency change rate is suppressed to a certain extent, and a stable state is recovered quickly.

Abstract: Provided is a neural network-based control method for switching a conduction mode of a Buck converter, including: establishing a state-space averaging model of the Buck converter, acquiring and substituting output voltages and inductive currents of the Buck converter in a discontinuous conduction mode (DCM) and a continuous conduction mode (CCM) into a state equation, and obtaining a precise state model under different working conditions; performing an explicit model predictive control (EMPC) design to generate a visual control law distribution diagram; obtaining neural network output through forward propagation, and comparing the neural network output with an optimal duty cycle obtained through EMPC, to obtain an error; performing fitting on the EMPC design, and completing offline training; and extracting a parameter of the trained neural network to a field programmable gate array (FPGA), and performing weighted summation calculation on input data, to obtain a duty cycle required for controlling the Buck con

Abstract: An intelligent garbage sorting robot includes a robot body and a walking base. The robot body is installed on the walking base, and a tool end of the robot body is installed with an executing mechanism. The executing mechanism includes a housing. An electromagnet is slidably disposed in the housing, the electromagnet is connected to a first driving assembly, the first driving assembly is configured to drive the electromagnet to slide, and an end of the housing is fixedly connected to multiple steel needles. Two mechanical claws symmetrically disposed on the walking base, and the two mechanical claws are spaced apart on the walking base. The two mechanical claws are respectively connected to second driving assemblies, and the second driving assemblies are configured to drive the two mechanical claws to move relatively.

Abstract: An intelligent garbage sorting robot is provided. The intelligent garbage sorting robot includes a robot body and a walking base. The robot body is installed on the walking base, and a tool end of the robot body is installed with an executing mechanism. The executing mechanism includes a housing. An electromagnet is slidably disposed in the housing, the electromagnet is connected to a first driving assembly, the first driving assembly is configured to drive the electromagnet to slide, and an end of the housing is fixedly connected to multiple steel needles. Two mechanical claws symmetrically disposed on the walking base, and the two mechanical claws are spaced apart on the walking base. The two mechanical claws are respectively connected to second driving assemblies, and the second driving assemblies are configured to drive the two mechanical claws to move relatively.

Abstract: The method includes: determining required virtual inertia based on a frequency change rate monitored at a grid connection point of a photovoltaic grid-connected system; obtaining values of indexes that affect virtual inertia provided by each virtual synchronous generator (VSG) control unit in a photovoltaic cluster of the photovoltaic grid-connected system; determining, by using a predetermined function equation, a function value corresponding to each index; and determining an allocation proportion of the required virtual inertia in each VSG control unit based on the function value corresponding to each index. A problem of difficult conversion between a plurality of kinds of performance during operation of a power system is resolved, the frequency change rate is suppressed to a certain extent, and a stable state is recovered quickly.

Abstract: A method and a system for controlling a gear of the two-speed gearbox and a vehicle is provided. The method for controlling a gear of a two-speed gearbox includes: determining a first target gear of the gearbox reflecting a current driving demand of a user according to a current vehicle-speed; and performing a gear intervention control of the gearbox based on at least one of the following aspects: position information of a drive gear-lever, a rotation-speed of a drive motor of a vehicle, and gear fault information of the gearbox, to shift the first target gear of the gearbox to a second target gear of the gearbox that matches with real-time operation conditions of the vehicle.

Abstract: A touch display device includes a backlight module (2) comprising a backlight iron frame (22); and a liquid crystal touch display panel (1) arranged on a light-emitting side of the backlight module; the liquid crystal touch display panel includes a touch electrode (11); and a compensation electrode (R) arranged on a side of the touch electrode facing the backlight module. The compensation electrode and the backlight iron frame are electrically connected with a ground signal terminal; and the touch electrode and the compensation electrode constitute a capacitor (C), a product of a capacitance value of the capacitor and a resistance value of the compensation electrode is smaller than a pulse width of a touch signal applied on the touch electrode, and an order of magnitude of the product is smaller than an order of magnitude of the pulse width of the touch signal.

Abstract: Disclosed are an array substrate, a manufacturing method thereof, and a display device. The array substrate includes: a base substrate; an electrode layer, located on the base substrate and including a first electrode and a second electrode which are spaced apart from each other; a touch electrode, located on the base substrate, and a touch wire, connected to the touch electrode; the touch electrode and the first electrode are insulated from each other, the touch electrode and the second electrode are insulated from each other, the touch wire is at one side of the touch electrode away from the base substrate, and the electrode layer is at one side of the touch electrode close to the base substrate.

Abstract: A touch display device includes a backlight module (2) comprising a backlight iron frame (22); and a liquid crystal touch display panel (1) arranged on a light-emitting side of the backlight module; the liquid crystal touch display panel includes a touch electrode (11); and a compensation electrode (R) arranged on a side of the touch electrode facing the backlight module. The compensation electrode and the backlight iron frame are electrically connected with a ground signal terminal; and the touch electrode and the compensation electrode constitute a capacitor (C), a product of a capacitance value of the capacitor and a resistance value of the compensation electrode is smaller than a pulse width of a touch signal applied on the touch electrode, and an order of magnitude of the product is smaller than an order of magnitude of the pulse width of the touch signal.

Abstract: A method and a system for controlling a gear of the two-speed gearbox and a vehicle is provided. The method for controlling a gear of a two-speed gearbox includes: determining a first target gear of the gearbox reflecting a current driving demand of a user according to a current vehicle-speed; and performing a gear intervention control of the gearbox based on at least one of the following aspects: position information of a drive gear-lever, a rotation-speed of a drive motor of a vehicle, and gear fault information of the gearbox, to shift the first target gear of the gearbox to a second target gear of the gearbox that matches with real-time operation conditions of the vehicle.

Abstract: An optical assembly and a virtual reality device are provided. The optical assembly includes a light-transmitting device, a light-converging device, and a light-reflecting device. The light-transmitting device is configured to receive light of a left-eye display image and light of a right-eye display image, and transmit the light of the left-eye display image and the light of the right-eye display image to a light-converging device. The light-converging device is configured to converge the light of the left-eye display image and the light of the right-eye display image to a light-reflecting device. A size of a first display image formed by converging the light of the left-eye display image is smaller than a size of the left-eye display image, and a size of a second display image formed by converging the light of the right-eye display image is smaller than a size of the right-eye display image.

Abstract: The fully-automatic closed-loop detection method includes: comparing a SCD file of a to-be-tested substation with a device-type data template file, so as to determine whether configuration information about the to-be-tested substation is correct; when the configuration information about the to-be-tested substation is correct, parsing the SCD file of the to-be-tested substation and generating a SSD topological diagram of the to-be-tested substation; and acquiring a testing item from a predetermined testing item library in accordance with the SSD topological diagram of the to-be-tested substation, generating a testing scheme for the to-be-tested substation, performing a testing operation and outputting a testing result.

Abstract: A flexible liquid-crystal-screen module and a method of assembling the flexible liquid-crystal-screen module are provided. The flexible liquid-crystal-screen module includes a groove-type middle bezel, a flexible liquid-crystal display and a backlight assembly, wherein, the backlight assembly is fixed at a bottom of a groove of the groove-type middle bezel, the flexible liquid-crystal display is fixed on an opening of the groove of the groove-type middle bezel, and the backlight assembly is spaced from the flexible liquid-crystal display.

Abstract: A display panel has a display area; and a peripheral area surrounding the display area, the peripheral area comprising a first area, a second area, a third area opposite to the first area, and a fourth area opposite to the second area. The display panel may further include a data driving integrated circuit provided in the first area; a gate drive circuit provided in the third area; and a plurality of signal transmission lines arranged in at least one of the second area and the fourth area. The configuration of the data driving integrated circuit and the gate drive circuit makes it possible to manufacture a display panel with a narrow bezel.