On-satellite high-reliability artificial intelligence microsystem electronic device

Abstract

The invention discloses an on-satellite high-reliability artificial intelligence microsystem electronic device which comprises a processing system, a storage system, a signal interface and a high-reliability reinforcing system. The processing system is used for receiving external data through a signal interface, calculating and processing satellite earth observation data information and storing calculation and processing results into the storage system; the storage system is used for storing data of the processing system; the signal interface is used for receiving external data and transmitting the data to the processing system; and the high-reliability reinforcement system is used for carrying out fault tolerance and fault protection on a processing system, a storage system and a signal interface.

Description

Technical Field

[0001] The present invention relates to a high-reliability on-board artificial intelligence microsystem electronic device, belonging to the field of aerospace technology. Background Art

[0002] At present, spacecraft in the fields of deep space exploration and remote sensing have high requirements for the intelligence and networking of information systems. Taking deep space exploration missions as an example, due to the long distance and the difficulty of real-time measurement and control on the ground, the detectors are required to have strong autonomous management and information processing capabilities, and the large amount of image data transmitted back to the earth poses a great challenge to the data transmission capability. Artificial intelligence microsystems are the carrier and foundation for realizing tasks such as spacecraft information extraction, environmental perception, autonomous judgment, and autonomous mission planning.

[0003] The corresponding AI chip products have been developed in the commercial field and have achieved certain results. However, in the field of aerospace, AI chips in the commercial field cannot be directly applied to spacecraft due to certain requirements for reliability, radiation resistance, efficiency, integration, applicability, and versatility. Therefore, it is necessary to carry out research on high-reliability, lightweight, and radiation-resistant architecture design methods based on COTS AI chips to realize microsystem electronic products that meet the special requirements of the aerospace field.

[0004] Traditional microsystem electronic devices. The main problems are as follows: (1) There is a problem of lack of carriers and foundations for onboard intelligent task processing. The present invention breaks through the intelligent lightweight microsystem architecture design method to form a highly reliable lightweight artificial intelligence microsystem suitable for onboard tasks. (2) There is a problem of insufficient radiation resistance of new artificial intelligence microsystems. The present invention aims at the fact that the radiation resistance indicators of existing COTS artificial intelligence chips cannot meet the needs of aerospace applications, breaks through the microsystem application protection design method, and meets the aerospace environment applicability requirements of the microsystem. (3) There is a problem of low versatility of traditional microsystem electronic products. The present invention breaks through the artificial intelligence architecture design method to meet the mission requirements of multiple aerospace models. Summary of the invention

[0005] The technical problem solved by the present invention is: to overcome the deficiencies of the prior art and provide a high-reliability on-board artificial intelligence microsystem electronic device that can cover on-board high-performance computing applications with high reliability and strong versatility.

[0006] The technical solution of the present invention is: an on-board high-reliability artificial intelligence microsystem electronic device, comprising: a processing system, a storage system, a signal interface and a high-reliability reinforcement system;

[0007] The processing system receives external data through the signal interface, calculates and processes the onboard earth observation data information, and stores the calculation and processing results in the storage system;

[0008] A storage system for storing data of the processing system;

[0009] A signal interface, used to receive external data and transmit the data to a processing system;

[0010] High-reliability reinforcement system, used for fault tolerance and failure protection of processing systems, storage systems, and signal interfaces.

[0011] The processing system includes a general-purpose processor CPU, a neural network processor NPU and a graphics processor GPU; the general-purpose processor CPU and the neural network processor NPU are interconnected through Gigabit Ethernet PHY; the general-purpose processor CPU is connected to the graphics processor GPU through the high-speed innovative computer expansion bus standard PCIe; the general-purpose processor CPU is used to control the graphics processor GPU, the neural network processor NPU and the high-speed innovative computer expansion bus standard PCIe, and send transmission and work instructions; the neural network processor NPU is used to perform deep learning calculations and reasoning of earth observation data, and send the calculation and reasoning results to the high-speed innovative computer expansion bus standard PCIe; the graphics processor GPU is used to compress and process earth observation images, and send the compression and processing results to the high-speed innovative computer expansion bus standard PCIe.

[0012] The operating main frequency of the general-purpose processor CPU is 2GHz; the neural network processor NPU has 4K high-definition video encoding and decoding capabilities, neural network reasoning computing capabilities and real-time encryption and decryption capabilities; the computing power of the graphics processor GPU is not less than 10Tops.

[0013] The storage system includes Flash and DDR storage systems; Flash is used to store instructions and control information sent by a general-purpose processor CPU; the DDR storage system is used to store graphics processor GPU and neural network processor NPU data transmitted via the high-speed innovative computer expansion bus standard PCIe.

[0014] The Flash capacity is at least 512M; the DDR storage system is of the fourth generation or above, with a capacity greater than 8Gb.

[0015] The signal interface includes an Ethernet interface, a mipi interface, a USB interface, an HDMI interface and a serial port.

[0016] The high-reliability reinforcement system includes a watchdog circuit and a triple-mode redundant radiation-resistant reinforcement circuit based on key instructions;

[0017] The watchdog circuit is used to monitor the functional errors caused by single-particle upsets inside the electronic device, and can automatically send a reset instruction if a functional abnormality occurs;

[0018] The triple-module redundant anti-radiation hardening circuit based on key instructions adopts a method of repeated execution of key instructions in the program, that is, the program is hardened by code redundancy; at the same time, triple-module redundant processing is performed on the key instructions, and only when the two judgments are consistent, the judgment result is considered correct.

[0019] Compared with the prior art, the present invention has the advantages that:

[0020] 1) This invention breaks through the intelligent lightweight microsystem architecture design method to form a highly reliable lightweight artificial intelligence microsystem suitable for onboard missions;

[0021] 2) The present invention breaks through the microsystem application protection design method and meets the aerospace environment applicability requirements of the microsystem;

[0022] 3) The present invention breaks through the artificial intelligence architecture design method to meet the mission requirements of multiple aerospace models. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 This is a block diagram of the device of the present invention.

[0024] Figure 2 This is the triple-module redundancy design diagram for key instructions. DETAILED DESCRIPTION

[0025] like Figure 1 As shown, the present invention provides a high-reliability on-board artificial intelligence microsystem electronic device, comprising:

[0026] Processing system: Artificial intelligence micro-module core processing unit system, including a high-performance general-purpose processor CPU, 2 NPU chips with neural network reasoning computing capabilities and real-time encryption and decryption capabilities, and a high-performance graphics processor GPU.

[0027] Storage system: including Flash and DDR storage systems.

[0028] Power supply system: including DC-DC chip and its power supply channel design.

[0029] Signal interface: including Ethernet interface, mipi interface, USB interface, HDMI interface and serial port.

[0030] High-reliability reinforcement system: including watchdog circuit and triple-mode redundant radiation-resistant reinforcement circuit based on key instructions.

[0031] The parameter requirements of the processing system include: a high-performance general-purpose processor FT-D2000 type CPU with a working main frequency of 2 GHz, two TS0845 type NPU chips with 4K high-definition video encoding and decoding capabilities, powerful neural network inference computing capabilities, and real-time encryption and decryption capabilities, and a high-performance graphics processor GPU with a computing power of not less than 10 Tops. The CPU and the NPU are interconnected through a gigabit Ethernet phy, and the CPU is connected to the GPU through PCIe.

[0032] The parameter requirements of the storage system include: at least 512M capacity Flash; a DDR memory of the fourth generation or above with a capacity greater than 8Gb.

[0033] The parameter requirements of the signal interface include: the DC-DC chip needs to cover voltages of 0.9V, 1.2V, 2.5V, and 3.3V; the NPU, GPU, and CPU are powered separately.

[0034] The parameter requirements of the signal interface include an Ethernet interface, a mipi interface, a USB interface, an HDMI interface, and a serial port.

[0035] The parameter requirements of the high-reliability reinforcement system include a watchdog circuit, which can monitor the functional errors caused by single-event upsets inside the electronic device. If a functional abnormality occurs, a reset instruction can be automatically sent. Based on the triple modular redundancy radiation-hardened design for critical instructions, for the critical instructions in the program, a method of repeated execution is adopted, that is, the program is hardened through code redundancy. Based on the time-filtering redundancy protection method for critical instructions, it is actually a triple modular redundancy processing for critical instructions. Only when the two judgments are consistent is the judgment result considered correct.

[0036] Through program instruction module division and module resource identification, the error robustness factor and resource overhead of each module without protection can be obtained.

[0037] The critical instructions refer to the instructions related to the overall process of the program. For example, conditional judgment instructions will directly determine the direction of the system process and affect the operation of the entire system. For such critical instructions in the program, a method of repeated execution is adopted, that is, the program is hardened through code redundancy. Based on the time-filtering redundancy protection circuit for critical instructions, it is actually a triple modular redundancy processing for critical instructions. Only when the two judgments are consistent is the judgment result considered correct. To illustrate the design idea of the triple modular redundancy design for critical instructions, taking the conditional judgment instruction (a < b) as an example, its design flow chart is as Figure 2 shown.

[0038] The error robustness factor is recorded as SERF, dimensionless; the resource overhead code size is recorded as C0, and its unit is byte; the data size is recorded as D0, and its unit is byte; the execution time is recorded as T0, and its unit is instruction cycle. k , which can be obtained by analyzing the code.

[0039] The error rates for key instructions are:

[0040] P2≈R p ×SERF×N k ×32

[0041] After the critical instructions are protected, the error rate after protection is:

[0042]

[0043] By comparing P3 and P2, it can be seen that the error probability is significantly reduced.

[0044] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.

Claims

1. A high-reliability on-board artificial intelligence microsystem electronic device, characterized in that: include: Processing system, storage system, signal interface and high-reliability reinforcement system; The processing system receives external data through the signal interface, calculates and processes the onboard earth observation data information, and stores the calculation and processing results in the storage system; A storage system for storing data of the processing system; A signal interface, used to receive external data and transmit the data to a processing system; High-reliability reinforcement system, used for fault tolerance and failure protection of processing systems, storage systems, and signal interfaces.

2. The on-board high-reliability artificial intelligence microsystem electronic device according to claim 1, characterized in that: The processing system includes a general-purpose processor CPU, a neural network processor NPU and a graphics processor GPU; the general-purpose processor CPU and the neural network processor NPU are interconnected through a gigabit Ethernet phy; the general-purpose processor CPU is connected to the graphics processor GPU through a high-speed innovative computer expansion bus standard PCIe; the general-purpose processor CPU is used to control the graphics processor GPU, the neural network processor NPU and the high-speed innovative computer expansion bus standard PCIe, and send transmission and work instructions; the neural network processor NPU is used to perform deep learning calculations and reasoning of earth observation data, and send the calculation and reasoning results to the high-speed innovative computer expansion bus standard PCIe; The graphics processing unit GPU is used to compress and process earth observation images, and send the compression and processing results to the high-speed innovative computer expansion bus standard PCIe.

3. The on-board high-reliability artificial intelligence microsystem electronic device according to claim 2, characterized in that: The operating main frequency of the general-purpose processor CPU is 2GHz; the neural network processor NPU has 4K high-definition video encoding and decoding capabilities, neural network reasoning computing capabilities and real-time encryption and decryption capabilities; the computing power of the graphics processor GPU is not less than 10Tops.

4. The on-board high-reliability artificial intelligence microsystem electronic device according to claim 1, characterized in that: The storage system includes Flash and DDR storage systems; Flash is used to store instructions and control information sent by a general-purpose processor CPU; the DDR storage system is used to store graphics processor GPU and neural network processor NPU data transmitted via the high-speed innovative computer expansion bus standard PCIe.

5. The on-board high-reliability artificial intelligence microsystem electronic device according to claim 4, characterized in that: The Flash capacity is at least 512M; the DDR storage system is of the fourth generation or above, with a capacity greater than 8Gb.

6. The on-board high-reliability artificial intelligence microsystem electronic device according to claim 1, characterized in that: The signal interface includes an Ethernet interface, a mipi interface, a USB interface, an HDMI interface and a serial port.

7. The on-board high-reliability artificial intelligence microsystem electronic device according to claim 1, characterized in that: The high-reliability reinforcement system includes a watchdog circuit and a triple-mode redundant radiation-resistant reinforcement circuit based on key instructions; The watchdog circuit is used to monitor the functional errors caused by single-particle upsets inside the electronic device, and can automatically send a reset instruction if a functional abnormality occurs; The triple-module redundant anti-radiation hardening circuit based on key instructions adopts a method of repeated execution of key instructions in the program, that is, the program is hardened by code redundancy; at the same time, triple-module redundant processing is performed on the key instructions, and only when the two judgments are consistent, the judgment result is considered correct.

Images