Space-based edge intelligent acceleration processing system

By using dual heterogeneous AI modules and FPGA intelligent control units in space-based edge devices, combined with domestic NPU and GPGPU chips, the problem of insufficient data processing capabilities and intelligence of satellite-based edge devices is solved, and efficient data analysis and The ability to make independent decisions improves communication bandwidth efficiency and autonomous control capabilities.

CN118964280AActive Publication Date: 2024-11-15NO 15 INST OF CHINA ELECTRONICS TECH GRP
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202411239801.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-09-05
Publication Date
2024-11-15
Estimated Expiration
2044-09-05

AI Technical Summary

Technical Problem

The prior art is difficult to effectively utilize AI chips to improve the data processing capabilities and intelligence of satellite-borne edge devices, resulting in low communication bandwidth efficiency and insufficient autonomous control capabilities.

Method used

Design a space-based edge intelligent acceleration processing system, adopting dual heterogeneous AI modules, intelligent management units, health management units and power supply units, combining domestic NPUs and GPGPU chips, intelligent management and task distribution through FPGAs, and achieve rapid data Handle and make decisions independently.

Benefits of technology

It significantly improves the data analysis and processing capabilities of space-based edge devices, enhances the level of independent decision-making and intelligence, supports a variety of deep learning frameworks and algorithm models, and improves load balancing and data privacy and security.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118964280A_ABST
    Figure CN118964280A_ABST
Patent Text Reader

Abstract

The invention, which relates to the technical field of the space-based communication equipment, discloses a space-based edge intelligent acceleration processing system comprising a dual-heterogeneous AI module, an intelligent management and control unit, a health management unit and a power supply unit. The dual-heterogeneous AI module comprises a first AI acceleration processing unit and a second AI acceleration processing unit which are respectively connected with the power supply unit; the intelligent management and control unit is connected with the first AI acceleration processing unit, the second AI acceleration processing unit, the health management unit and the power supply unit. The health management unit is connected with the first AI acceleration processing unit, the second AI acceleration processing unit and the power supply unit. The AI chip is applied to space-based edge intelligent acceleration calculation, space-based intelligent calculation, space-based edge computing power balance and hardware support of endogenous AI are achieved, and the on-orbit data analysis and processing and edge autonomy capabilities of space-based edge equipment are greatly improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of space-based communication equipment, and more specifically to a space-based edge intelligent acceleration processing system. Background Art

[0002] With the substantial improvement of satellite performance and the data explosion brought by large-scale constellations, satellite data transmission and satellite technology applications are facing problems such as insufficient communication bandwidth between satellites and the ground, limited ground base station resources and coverage. At the same time, as satellite missions become more and more complex, there are often emergencies and changes in the operating environment. Therefore, intelligent acceleration technology is used on satellite-borne edge devices to extract key information and filter redundant information. While improving the efficiency of communication bandwidth, it further improves the autonomous control and decision-making capabilities of the edge end, and promotes the intelligent development of satellite control and scheduling.

[0003] In the past two years, ground-based artificial intelligence has developed explosively, comprehensively promoting the multi-ecological integration of AI and technologies in various industries. A global trend of AI leading technological changes in various industries has emerged. At the same time, AI technology, industry, and applications have ushered in a breakthrough period of rapid iteration and exploration. my country has also made rapid progress in the field of AI technology, especially domestic high-performance dedicated AI chips, which have achieved breakthrough development and wide application in the field of intelligent accelerated computing through their excellent computing power, optimized algorithms, and powerful data models.

[0004] Therefore, how to apply AI chips to the intelligent acceleration of satellite-borne edge devices, so as to further improve the data processing capabilities and intelligence level of satellite-borne edge devices, is an issue that technical personnel in this field urgently need to solve. Summary of the invention

[0005] In view of this, the present invention provides a space-based edge intelligent acceleration processing system, which applies AI chips to space-based edge intelligent acceleration computing to meet the needs of large-scale computing power and big data models in complex task processing of low-Earth orbit satellites, multi-satellite collaboration in emergency situations, intelligent scheduling of distributed edge computing node tasks, and intelligent management and control of satellite-borne services. It realizes space-based intelligent computing, space-based edge computing power balance, and hardware support for endogenous AI, greatly improving the on-orbit data analysis, processing and edge autonomy capabilities of space-based edge devices.

[0006] In order to achieve the above object, the present invention adopts the following technical solution:

[0007] A space-based edge intelligent acceleration processing system, comprising: a dual-heterogeneous AI module, an intelligent management and control unit, a health management unit and a power supply unit; the dual-heterogeneous AI module comprises a first AI acceleration processing unit and a second AI acceleration processing unit, which are respectively connected to the power supply unit; the intelligent management and control unit is respectively connected to the first AI acceleration processing unit, the second AI acceleration processing unit, the health management unit and the power supply unit; the health management unit is respectively connected to the first AI acceleration processing unit, the second AI acceleration processing unit and the power supply unit.

[0008] Preferably, the first AI acceleration processing unit includes an NPU chip, a first SPI Flash circuit and a memory circuit. The first SPI Flash circuit uses a board-mounted Flash chip mounted on the SPI bus of the NPU chip to store NPU chip-related configuration information. The memory circuit uses board-mounted LPDDR4 particles to be connected to a 96-bits LPDDR4X controller integrated inside the NPU chip for storing processing data, supporting a maximum of 12GB; the second AI acceleration processing unit includes a GPGPU chip and a second SPI Flash circuit. The second SPI Flash circuit uses a board-mounted Flash chip mounted on the SPI bus of the GPGPU chip to store GPGPU chip-related configuration information. The GPGPU chip itself is sealed with an HBM2E memory with a bandwidth support of 800GB / s and a capacity of 32GB for storing processing data; the NPU chip and the GPGPU chip are connected to the intelligent management and control unit using a high-speed data bus.

[0009] Preferably, the intelligent management and control unit includes an FPGA chip, a configuration circuit, an information storage circuit, a PCIe interface circuit, a GPIO interface and an I 2 C interface; the configuration circuit is directly connected to the FPGA chip using a board-mounted NOR Flash chip; the information storage circuit uses a board-mounted DDR3 storage particle to connect to the DDR3 memory controller inside the FPGA chip; the PCIe interface circuit is connected to the PCI bus interface of the FPGA chip; the GPIO interface is connected to the FPGA chip, and is respectively connected to the GPIO interfaces of the NPU chip and the GPGPU chip; the I 2 The C interface is connected to the FPGA chip and the health management unit respectively. The on-board mission data is identified and judged by the logic unit inside the FPGA chip, and the on-board mission data is classified into low-level data and high-level data. The low-level data is transmitted to the NPU chip for processing, and the high-level data is transmitted to the GPGPU chip for processing; the low-level data includes image recognition mission data, intelligent optimization mission data, etc., and the high-level data includes target recognition mission data, planning mission data, intelligent control mission data, etc.

[0010] Preferably, the FPGA is connected to the space-based edge device, the NPU chip and the GPGPU chip respectively through the PCIe bus, and is connected to the NPU chip and the GPGPU chip respectively through the GPIO interface, so as to intelligently control the acceleration processing system; the PCIe bus is used to realize data transmission with the space-based edge device, and is used to realize the distribution of different levels of task data on the NPU chip and the GPGPU chip and the feedback of the task acceleration results generated by the NPU chip and the GPGPU chip; the GPIO interface is used to realize the online status recognition and task assignment function of the intelligent control unit for the dual heterogeneous AI module. The GPIO interface can identify and feedback whether the NPU chip or the GPGPU chip has the status of processing the task, and pull up or lower a pin of the NPU chip or the GPGPU chip through the GPIO interface, so that the FPGA chip can read the working status of the two chips later and realize task assignment.

[0011] Preferably, the health management unit includes an embedded MCU and several temperature sensors, the temperature sensors are connected to the embedded MCU, and the embedded MCU is connected to the embedded MCU through I 2 The I2C bus is connected to the I2C bus of the intelligent control unit. 2 C interface (to realize connection with FPGA chip), the NPU chip and the GPGPU chip, the embedded MCU is connected to the space-based edge device through the IPMB bus; the temperature sensor collects temperature and transmits it to the embedded MCU, and the ADC converter integrated in the embedded MCU collects voltage.

[0012] Preferably, an anti-radiation protective cover based on a tantalum-tungsten-aluminum composite material is provided on the outside of the dual-heterogeneous AI module.

[0013] It can be seen from the above technical solution that compared with the prior art, the present invention discloses a space-based edge intelligent acceleration processing system, which is based on a dedicated AI chip with two heterogeneous computing resources, namely, domestic NPU and GPGPU. Through the enhanced processing of on-board video, image and other information by high-performance domestic FGPA and the management and control of the heterogeneous computing power of dual dedicated AI chips, the space-based edge device realizes the acceleration capability of intelligent computing when processing massive video and image tasks and the autonomous decision-making capability when coping with complex tasks and emergency situations. In addition, the health information management function of the intelligent acceleration processing module is realized through the high-reliability MCU, which provides hardware support for the visualization of the status of equipment between the satellite and the ground. The power supply unit provides the required power supply voltage for the units of the space-based edge intelligent acceleration processing system. Among them, the intelligent acceleration processing architecture based on domestic NPU+GPGPU dual heterogeneous computing power resources in the present invention supports a variety of deep learning frameworks, has a wide coverage of deep learning algorithm models, has continuously tuned operator units and acceleration components, and supports the rapid iteration of various application algorithms on the satellite; the bidirectional data transmission mode of FPGA+heterogeneous AI can realize the intelligent management and control and task distribution of dual heterogeneous computing power resources without switching chips, and directly carry high-speed data buses. FPGA performs intelligent scheduling according to the complexity of the task, further accelerating the timeliness of space-based edge devices to process complex tasks, and improving the load balancing ability of space-based edge devices. In summary, the present invention is aimed at AI computing in the application scenario of space-based edge nodes. Through the special architecture and algorithm optimization of the intelligent acceleration processing module, the timeliness of the on-board computer to analyze and process massive videos, pictures and other information in space in real time is accelerated, and the satellite's ability to calculate and plan tasks autonomously in orbit is improved, helping the satellite constellation to develop into an independent operating body that can be controlled and managed autonomously, and promoting the intelligent development of space-based communications. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0015] Figure 1 A schematic diagram of the structure of the space-based edge intelligent acceleration processing system provided by the present invention;

[0016] Figure 2 A schematic diagram of the data flow control topology of the space-based edge intelligent acceleration processing system provided by the present invention;

[0017] Figure 3 A flow chart of heterogeneous resource control and information preprocessing unit-data processing provided by the present invention;

[0018] Figure 4 This is a block diagram of the power supply principle of the space-based edge intelligent acceleration processing system provided by the present invention. DETAILED DESCRIPTION

[0019] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0020] The embodiment of the present invention discloses a space-based edge intelligent acceleration processing system, the structure of which is as follows: Figure 1 As shown, it includes: a dual-heterogeneous AI module, an intelligent management and control unit, a health management unit and a power supply unit; the dual-heterogeneous AI module includes a first AI acceleration processing unit and a second AI acceleration processing unit, which are respectively connected to the power supply unit; the intelligent management and control unit is respectively connected to the first AI acceleration processing unit, the second AI acceleration processing unit, the health management unit and the power supply unit; the health management unit is respectively connected to the first AI acceleration processing unit, the second AI acceleration processing unit and the power supply unit.

[0021] Furthermore, the first AI acceleration processing unit adopts NPU chip, and the second AI acceleration processing unit adopts GPGPU chip. The NPU chip and GPGPU chip are connected to the intelligent control unit using a high-speed data bus. Domestic NPU and domestic GPGPU dedicated AI chips, as the core processing units of the space-based edge intelligent acceleration processing module, provide space-based edge devices with heterogeneous computing resources for different task requirements. With the big data model, deep learning acceleration operator library, acceleration tool chain and other components and optimized intelligent algorithms integrated in the two dedicated AI chips, the space-based edge devices are improved in processing massive videos, pictures and other complex intelligent computing tasks. The processing time efficiency of space-based edge devices in AI computing application scenarios is accelerated, so that space-based edge devices have certain edge autonomy capabilities when handling emergencies, and can realize intelligent task scheduling and autonomous decision-making, etc., providing hardware support for space-based edge devices to realize satellite-ground collaborative scheduling algorithms, intelligent routing, and multi-dimensional fusion intelligent communication applications of the sky and the earth. At the same time, with the strong edge computing capabilities of dedicated AI chips, space-based edge devices have higher data privacy and security, especially suitable for the application of space-based edge nodes.

[0022] Furthermore, the first AI acceleration processing unit includes an NPU chip, a first SPI Flash circuit and a memory circuit. The first SPI Flash circuit uses a board-mounted Flash chip mounted on the SPI bus of the NPU chip to store relevant configuration information of the NPU chip. The memory circuit uses board-mounted LPDDR4 particles to be connected to the 96-bits LPDDR4X controller integrated inside the NPU chip for storing processing data, supporting a maximum of 12GB; the second AI acceleration processing unit includes a GPGPU chip and a second SPI Flash circuit. The second SPI Flash circuit uses a board-mounted Flash chip mounted on the SPI bus of the GPGPU chip to store relevant configuration information of the GPGPU chip. The GPGPU chip itself is sealed with an HBM2E memory with a bandwidth support of 800GB / s and a capacity of 32GB for storing processing data.

[0023] Furthermore, the dual-heterogeneous AI module uses the NPU chip model Ascend 310B and the GPGPU chip model Zhikai 100. These two heterogeneous dedicated AI chips serve as the core processing devices of the space-based edge intelligent acceleration processing system, helping the intelligent acceleration processing system and even the space-based edge devices to realize the AI ​​acceleration processing functions required by different tasks, and have the ability to respond to emergency situations, environmental changes, and other tasks at any time. Among them, the NPU chip has a computing power of up to 20TOPS, which can perform complex image recognition, intelligent optimization and other tasks; the GPGPU chip has a computing power of up to 192TOPS, which can perform complex target recognition, task planning, intelligent control and other tasks. Therefore, the two AI acceleration processing units operate independently with different divisions of labor. Through their own architectural advantages, big data models, deep learning acceleration operator libraries, acceleration tool chains, deep learning tuning engines and other operating environment components, they process appropriate tasks, reduce the number of core codes and have the characteristics of coping with operator diversity, providing a hardware foundation for the coordinated cooperation of heterogeneous computing power of space-based edge intelligent acceleration processing modules, and AI load balancing and intelligent scheduling of space-based edge devices.

[0024] Furthermore, the NPU chip supports 8Lane SerDes, which is distributed in 2 SerDes Macros. It can realize flexible configuration of GE, USB 3.0, PCIe (supporting EP / RC) and SATA according to the application scenarios of different products. As a slave device of the intelligent management and control unit, the chip needs to configure SerDes into the PCIe EP usage scenario. The SerDes configuration in NPU EP mode is shown in Table 1 below, and the PCIE_EP_RC_FLASH pin is configured to a high level to realize its ability as a slave device of the intelligent management and control unit, thereby completing the interaction of data information; the GPGPU chip provides a PCIe4.0x16 bus interface to the outside world, which is backward compatible and uses its x8 lane to directly connect to the PCIe bus of the intelligent management and control unit to realize the interaction of data information.

[0025] Table 1 NPU SerDes configuration (EP) mode

[0026]

[0027] Furthermore, the debugging interface of the dual-heterogeneous AI module is mainly UART and JTAG. The NPU chip has 7 UART interfaces, and the GPGPU chip has 2 UART interfaces and 1 JTAG interface. After the UART interface is converted through LevelShift, the online printing function based on RS422 level is realized; the JTAG interface can be directly brought out; the debugging interface of the dual-heterogeneous AI module is brought out through the front panel for easy debugging.

[0028] Furthermore, the intelligent control unit uses an FPGA chip; Figure 2 As shown, the FPGA chip is connected to the NPU chip and the GPGPU chip. The on-board mission data is received or reported through the PCIe bus of the FPGA chip, and the on-board mission data is identified and judged using its internal logic unit. The tasks of image recognition and intelligent optimization with relatively low complexity are assigned to the AI ​​acceleration processing unit based on the domestic NPU, and the tasks of target recognition, mission planning, intelligent control and other tasks with relatively high complexity are assigned to the AI ​​acceleration processing unit based on the domestic GPGPU, so as to realize the intelligent allocation of different task levels on the satellite. The intelligent control of the dual-heterogeneous computing power AI acceleration processing unit by FPGA is realized through the PCIe bus and GPIO interface respectively, where the PCIe bus is used to realize the distribution of different levels of on-board mission data and the return of task acceleration results for the dual-heterogeneous computing power AI acceleration processing unit; the GPIO interface is used to realize the online recognition and task assignment functions of the intelligent control unit for the dual-heterogeneous AI module.

[0029] Furthermore, the intelligent management and control unit adopts the domestic V7 series FPGA, which mainly realizes the enhanced pre-processing function of massive videos, pictures and other task information from the satellite, the intelligent allocation function for tasks of different complexity and the management and control function of heterogeneous computing resources.

[0030] First, the intelligent control unit receives massive AI task information to be processed from the space-based edge device through a high-speed bus. Second, relying on its high-speed parallel computing processing unit, it has high-performance digital signal processing capabilities and high-bandwidth data throughput capabilities to pre-process the video, image and other information to be processed, enhance the task information, reduce the load burden of the AI ​​computing unit, and improve the timeliness of the AI ​​computing unit. Third, through its rich programmable logic resources, the information to be processed is combined with the current on-board tasks to make intelligent judgments on the complexity of the task, and annotate it. According to the annotated task complexity, the information to be processed and the task are intelligently allocated, among which tasks with low task complexity suitable for image recognition, intelligent optimization, etc. are allocated to the AI ​​acceleration processing unit based on the domestic NPU, and tasks with high task complexity and high computing power requirements such as target recognition, task planning, intelligent control, etc., especially emergency tasks, are allocated to the AI ​​acceleration processing unit based on the domestic GPGPU, which improves the energy efficiency and utilization rate of the two independently running intelligent processing units while accelerating the computing timeliness of the space-based edge device. Finally, the data processed by the dual heterogeneous AI modules is recycled and packaged for processing and transmitted to the space-based edge device through a high-speed bus. This completes the preprocessing, control and distribution, and result feedback process of an AI computing task. 2 The C interface is directly connected to the health management unit, which reads its health information.

[0031] Among them, preprocessing refers to first converting the image data into a digital signal using the ADC converter inside the FPGA chip, then using various filter algorithms inside the FPGA chip to filter the image, and using various edge algorithms inside the FPGA chip to perform edge detection, image enhancement and other processing. Finally, the digital signal is converted into an analog signal through the DAC converter inside the FGPA chip, and the FPGA chip transmits the analog signal to the NPU chip or GPGPU chip.

[0032] Furthermore, the intelligent management and control unit is provided with a configuration circuit, an information storage circuit, a PCIe interface circuit, a GPIO interface, an I 2C interface and debugging interface, etc.; Among them, the configuration circuit adopts the board-mounted NOR Flash chip, which is directly connected to the FPGA chip, and uses JTAG, AS and other configuration methods to download the required configuration program to the Flash to realize the configuration management of the FPGA. AS refers to the active serial mode. After the UART interface is converted through Level Shift, the online printing function based on the RS422 level is realized. The programming data between the logic blocks and interconnections of the FPGA is defined through the configuration circuit. When the FPGA is powered on, such data is loaded into the device to realize the ability of FGPA to execute the predetermined function; the information storage circuit adopts the board-mounted DDR3 storage particles to connect the DDR3 memory controller inside the FPGA chip to realize the high-speed data transmission and reception, caching and parallel processing functions of the intelligent management and control unit, and the maximum speed supports 1866Mb / s, with ECC error checking function; the PCIe interface circuit is connected to the 3 PCI bus interfaces of the FPGA chip, and the PCI bus interface supports PCIe at most 3.0x8, the AI ​​computing tasks of the space-based edge device enter the intelligent management and control unit through the PCIe interface circuit and PCIe bus interface, and are distributed to the dual heterogeneous AI modules through the PCIe bus after preprocessing and management distribution. Subsequently, the data information processed by the dual heterogeneous AI modules returns to the intelligent management and control unit, and after packaging processing, it is fed back to the space-based edge device; the GPIO interface is directly connected to the GPIO interface of the NPU chip and the GPGPU chip, and is mainly used to realize the online identification and task distribution of the dual heterogeneous AI modules. It relies on the very rich programmable logic units inside its core device FPGA, and can completely cover the GPIO interface of the dual heterogeneous AI module; I 2 C interface directly with the health management unit’s I 2 C interface connection, the health management unit realizes real-time collection of its working status and other information, realizes real-time monitoring and reset, power-on control of the working status of the intelligent management and control unit, etc.; the debugging interface is mainly USB interface and LVDS interface. Among them, the USB interface is directly configured by the FPGA chip output, so ESD protection needs to be added. In addition, the LVDS interface level of the FPGA chip is 1.8V, which is different from the standard LVDS interface 1.2V level. Therefore, Level Shift is required for level conversion to realize the display output function. The debugging interface is led to the front panel connector for easy debugging.

[0033] Furthermore, the health management unit includes an embedded MCU and several temperature sensors, the sensors are connected to the embedded MCU, and the embedded MCU 2The C bus connects the FPGA chip, NPU chip and GPGPU chip respectively, and connects to the space-based edge device through the IPMB bus; several temperature sensors collect temperature, and the ADC converter integrated in the embedded MCU collects voltage. Based on the domestically produced high-reliability embedded MCU, with onboard sensor devices, it monitors the key voltage, temperature and other information on the space-based edge intelligent acceleration processing system in real time. The health management unit reports its health status information to the space-based edge device through the IPMB bus, realizing the working status alarm function of the processing module. It has the control capabilities of module power-on, reset, etc., providing a hardware foundation for the overall health management of space-based edge devices. The NPU chip has 7 I 2 C interface, supports 1-way Slave, 6-way Master, and the GPGPU chip has 4 I 2 C interface, supports 1-way Slave, 3-way Master, NPU chip and GPGPU chip through I 2 The C bus is directly connected to the embedded MCU, and the health management unit reads the health information of the dual-heterogeneous AI modules, realizes real-time monitoring and reset of the working status of the dual-heterogeneous AI modules, and power-on timing control. The embedded MCU is also connected to the over-voltage, under-voltage and over-temperature protection modules, and sends a reset signal to the over-voltage, under-voltage and over-temperature protection modules to reset them.

[0034] Furthermore, the health management unit uses the domestically produced 32F4 series embedded MCU based on the ARM Cortex-M4 32bit RISC core. 2 The C bus realizes real-time monitoring of the working status of the space-based edge intelligent acceleration processing system. In addition, the voltage of key points is collected through the internal integrated ADC converter, and the collected voltage and temperature are converted into analog voltage, temperature parameters and other information. It also makes early warning judgments based on the set thresholds, generates early warning information and reports it to the space-based edge device, which plays the role of fault early warning. The monitoring units are monitored through their respective I 2 The C bus is directly mounted on the health management unit to complete the working status reporting and accept power-on, reset and other functional controls.

[0035] Furthermore, the embedded MCU has a complete minimum system circuit, including a clock circuit, a Flash storage circuit, and a BOOT startup and RST configuration circuit; the embedded MCU supports three clock sources, namely HIS, HSE, and main PLL, to drive the internal system clock SYSCLK, and has two secondary clock sources, 32kHz and 32.768kHz. In this embodiment, the HSE clock source (25MHz) of an external crystal oscillator with higher precision can be used to drive the ARM system clock (168MHz), and the various clocks required by the APB bus peripherals (such as ADC) are realized by the internal PLL multiplication and division. In addition, a 32.768kHz LSE crystal oscillator is used as the AR The clock source of the M real-time clock RTC makes it easy for the health management unit to record the system time in real time; the Flash storage circuit supports 128-bit wide data reading capability, which is realized by hanging the SPI data type Flash chip on the SPI bus interface of the embedded MCU, and is used for the health management unit to cache data information; the BOOT startup and RST configuration circuits are used to configure the BOOT1 and BOOT0 pins. The MCU chip startup mode is controlled by the BOOT1 and BOOT0 pins. By pulling down to set the 00 mode, the main Flash startup mode of the health management unit is realized; the MCU chip reset pin NRST is low and effective, and the on-board key switch realizes the reset function of the health management unit.

[0036] Furthermore, the health management unit has I 2 C interface, ADC functional circuit and corresponding debugging interface; I 2 The C interface is used to collect the working status of the dual heterogeneous AI modules and intelligent processing units, and to control the reset and power-on timing. The embedded MCU has three I 2 C bus interface, supports multiple uses, such as CRC generation and verification, SMBus and PMBus, etc., and also supports the use of general GPIO interface to simulate I 2 C protocol implementation 2C bus function; the ADC function circuit realizes the collection and conversion of analog quantities such as main voltage parameters and temperature information. The MCU chip extracts and processes the converted digital signals to realize real-time monitoring of the voltage and temperature of the space-based edge intelligent computing module. The embedded MCU has a 12-bit successive approximation ADC converter with up to 19 multiplexed channels, of which ADC[15:0]16 channels can be provided to the outside, ADC[13:0]14 channels are respectively connected to the voltage division processing circuit to collect the 12V total voltage, the main voltage of the NPU chip, the main voltage of the GPGPU chip, and the main voltage of the FPGA chip. ADC[15:14]2 channels collect the temperature information close to the temperature measurement point of the NPU chip and the temperature information of the temperature measurement point of the GPGPU chip. ADC16 is dedicated to collecting temperature information close to the temperature measurement point of the FPGA chip, and VREF is connected to the reference voltage. The A / D conversion of these channels supports single, continuous, scanning or discontinuous sampling modes. By setting the reference voltage (VREF = 2.5V) and voltage division processing, the analog quantity acquisition of 0 ~ VREF can be realized; the debugging interface is the RS422 interface and the SWD interface. The RS422 interface realizes the RS422 level after the Level Shift level conversion, supports online information printing, and the SWD interface is the SW debugging interface of the MCU. The program can be downloaded through the SWCLK and SWDIO dual lines. The debugging interface is connected to the front panel connector for easy debugging. The health management unit covers the voltage measurement points and temperature measurement points of all functional units, which is convenient for early warning and real-time monitoring of the working status of the space-based edge intelligent acceleration processing system.

[0037] Furthermore, the power supply unit provides corresponding power supply voltage for each functional unit of the space-based edge intelligent acceleration processing system, and adopts a universal and modular design method to ensure that the space-based edge intelligent acceleration processing system can operate stably and reliably.

[0038] Furthermore, domestically produced dedicated AI chips themselves do not have radiation resistance capabilities. Therefore, during the design and implementation, it is necessary to adopt certain protective measures such as anti-radiation shields based on tantalum-tungsten-aluminum composite materials so that they can be stably and reliably used in space environments.

[0039] On the other hand, in a specific embodiment, the intelligent management and control unit performs the following processing flow: Figure 3As shown, specifically: Step 1: Receive the data cache of the on-board task of the space-based edge device and perform enhanced processing; Step 2: Determine the task type of the enhanced data and divide the data into low-level tasks and high-level tasks; Step 3: Distribute the data contained in the low-level task to the NPU chip for processing to obtain the low-level task processing result, and distribute the data contained in the high-level task to the GPGPU chip for processing to obtain the high-level task processing result; Step 4: Feedback the low-level task processing results and the high-level task processing results to the space-based edge device.

[0040] On the other hand, in a specific embodiment, the structural frame of the power supply unit is as follows Figure 4 As shown, it includes three groups of power supply circuits and one group of current limiting protection circuits. Each group of power supply circuits includes a first isolated DC / DC module, a current limiting and over-voltage and under-voltage protection module, and a second isolated DC / DC module connected in sequence; the three groups of power supply circuits respectively connect the FPGA chips of the first AI acceleration processing unit, the second AI acceleration processing unit, and the intelligent management and control unit to a 12V DC power supply; the health management unit is connected to a 12V DC power supply through a current limiting protection circuit, and the health management unit is respectively connected to the second isolated DC / DC modules of the three groups of power supply circuits.

[0041] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0042] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A space-based edge intelligent acceleration processing system, characterized in that: include: Dual heterogeneous AI module, intelligent management and control unit, health management unit and power supply unit; the dual heterogeneous AI module includes a first AI acceleration processing unit and a second AI acceleration processing unit, which are respectively connected to the power supply unit; the intelligent management and control unit is respectively connected to the first AI acceleration processing unit, the second AI acceleration processing unit, the health management unit and the power supply unit; the health management unit is respectively connected to the first AI acceleration processing unit, the second AI acceleration processing unit and the power supply unit.

2. A space-based edge intelligent acceleration processing system according to claim 1, characterized in that: The first AI acceleration processing unit includes an NPU chip, a first SPI Flash circuit and a memory circuit. The first SPI Flash circuit uses a board-mounted Flash chip to be mounted on the SPI bus of the NPU chip, and the memory circuit uses board-mounted LPDDR4 particles to be connected to a 96-bits LPDDR4X controller integrated in the NPU chip; the second AI acceleration processing unit includes a GPGPU chip and a second SPI Flash circuit. The second SPI Flash circuit uses a board-mounted Flash chip to be mounted on the SPI bus of the GPGPU chip, and the NPU chip and the GPGPU chip are connected to the intelligent management and control unit using a high-speed data bus.

3. A space-based edge intelligent acceleration processing system according to claim 2, characterized in that: The intelligent management and control unit includes an FPGA chip, a configuration circuit, an information storage circuit, a PCIe interface circuit, a GPIO interface and an I 2 C interface; the configuration circuit is directly connected to the FPGA chip using a board-mounted NOR Flash chip; the information storage circuit is connected to the DDR3 memory controller inside the FPGA chip using a board-mounted DDR3 storage particle; the PCIe interface circuit is connected to the PCI bus interface of the FPGA chip; The GPIO interface is connected to the FPGA chip, and is respectively connected to the GPIO interfaces of the NPU chip and the GPGPU chip; 2 The C interface is connected to the FPGA chip and the health management unit respectively.

4. The space-based edge intelligent acceleration processing system according to claim 3, characterized in that: The FPGA chip is respectively connected to the space-based edge device, the NPU chip and the GPGPU chip through the PCIe bus.

5. The space-based edge intelligent acceleration processing system according to claim 3, characterized in that: The health management unit includes an embedded MCU and several temperature sensors, wherein the temperature sensors are connected to the embedded MCU. 2 The I2C bus is connected to the I2C bus of the intelligent control unit. 2 C interface, the NPU chip and the GPGPU chip, and the embedded MCU is connected to the space-based edge device through the IPMB bus.

6. The space-based edge intelligent acceleration processing system according to claim 1, characterized in that: An anti-radiation protective cover based on a tantalum-tungsten-aluminum composite material is arranged outside the dual-heterogeneous AI module.

Citation Information

Patent Citations

  • Space-based edge calculation module and equipment

    CN116501508A

  • Onboard AI-based Cloud Detection System for Enhanced Satellite Autonomy Using PUS

    US20220366684A1