Heterogeneous computing platform for satellite-borne health management and data interaction scheduling method

Abstract

The invention provides a heterogeneous computing platform for satellite-borne health management and a data interaction scheduling method, and belongs to the technical field of aerospace. The invention aims to solve the problems of control instruction blockage, rate mismatching between data receiving and model processing and incapability of adapting to a high-interference environment in a traditional satellite health management architecture. And the heterogeneous computing platform comprises a dual-Ethernet communication link and a satellite-borne health management single machine which are connected in sequence. The data interaction scheduling method comprises the following steps: S1, performing data encapsulation on a high-frequency telemetry data frame and a control instruction, and introducing a verification mechanism into a dual-Ethernet communication link; s2, uploading data and instructions through a communication link and a control link; s3, in the shared memory area, data writing and reading are carried out through a double-cache mechanism, and automatic correction and detection alarm are carried out on the stored data through an ECC error correction coding mechanism; and S4, performing link binding and priority ranking execution on the tasks according to the health management module.

Description

Technical Field

[0001] This invention relates to a heterogeneous computing platform and data interaction scheduling method for spaceborne health management, belonging to the field of aerospace technology. Background Technology

[0002] The rapid increase in the number of satellites in orbit and the continuous rise in mission complexity have led to enhanced coupling between various subsystems within satellites, making system operation more dynamic and unpredictable, and significantly increasing the probability of critical subsystem failures. Statistical data shows that core modules such as power systems, thermal control systems, and attitude control systems are all at risk of sudden or degenerative failures to varying degrees during their service life, which can lead to complete satellite malfunction in severe cases. Therefore, how to continuously and accurately monitor and assess the health status of satellites has become a crucial issue for ensuring mission reliability.

[0003] Traditional satellite health management architectures typically rely on ground stations periodically receiving telemetry data and using post-processing algorithms to analyze and diagnose the satellite's operational status. However, due to limitations in telemetry and control coverage caused by geographical location and the visible window of the satellite-to-ground link, there are significant intervals and delays in the transmission of telemetry data from the satellite to the ground, making continuous observation difficult. Furthermore, due to bandwidth and latency constraints in the satellite-to-ground transmission link, high-dimensional telemetry data (especially high-frequency, rapidly updating device-level status parameters) often cannot be transmitted completely and in real time. By the time the ground completes data decoding, anomaly detection, and response generation, the optimal handling window may have already been missed, leading to further deterioration of the fault and even triggering a chain reaction of failures.

[0004] To compensate for the real-time limitations of ground-based processing methods, the industry has gradually shifted towards an "on-board computing" model in recent years, migrating some or all health management functions to the on-orbit satellite platform for local execution. The essence of this strategy is to embed complex algorithms such as anomaly detection, trend prediction, and fault tree diagnosis within the satellite computing platform, enabling it to independently complete the closed-loop processing flow from data acquisition to fault alarm. This transformation not only improves the timeliness of health management but also reduces the load on telemetry data transmission, freeing up bandwidth for space-to-ground communication resources.

[0005] However, to truly realize the autonomous operation mode of "on-board health management", it is necessary to rely on the powerful on-board computing platform behind it. The on-board computing platform is facing more stringent interaction pressure. Typical problems include: (1) Data link reuse leads to control command blockage; telemetry data and control commands share a single link (such as Ethernet or serial port), which is very easy to cause congestion when data is injected under high load, resulting in command response delay or even link blockage, which cannot meet the real-time response requirements of "command priority and data concurrency" in on-board health management. (2) There is a caching bottleneck between high-speed data injection and model calculation; telemetry data has a high sampling frequency. If the traditional single buffer method is used, the platform processing rate is often lower than the injection rate, resulting in data overwriting, missing frames or processing lag, making it difficult to maintain the continuous operation of the model in high-concurrency scenarios. (3) Data transmission error problem in on-orbit interference environment; high-energy particles or electromagnetic radiation in space can easily cause bit flipping in memory or transmission channels. Traditional parity check is not enough to provide strong error correction capability, which can easily lead to model interpretation errors or false alarms. (4) Communication task scheduling mechanism lacks intelligent support. Traditional data links use a "first-come, first-served" model, which cannot intelligently allocate resources based on task urgency and communication channel load. This can easily lead to low-priority tasks occupying high-priority channels, causing scheduling errors and reduced communication efficiency. Summary of the Invention

[0006] To address the problems of control command congestion, data reception and model processing rate mismatch, and inability to adapt to high-interference environments in traditional satellite health management architectures, this invention proposes a heterogeneous computing platform and data interaction scheduling method for onboard health management.

[0007] The technical solution adopted by this invention to solve the above problems is: the heterogeneous computing platform for spaceborne health management proposed in this invention includes: Dual Ethernet communication links are used to receive high-frequency telemetry data frames from the ground monitoring station and transmit them to the onboard health management unit, receive control commands issued by the ground monitoring station and transmit them to the onboard health management unit, and transmit the processing results, abnormal information and output content of the onboard health management unit back to the ground monitoring station. The onboard health management unit is bidirectionally connected to the dual Ethernet communication links. The onboard health management unit adopts a dual buffering mechanism, an anti-interference transmission and verification mechanism, and a dual link priority scheduling mechanism to receive control class execution shortest path response and adaptively allocate the tasks to be processed to the idle links in the dual Ethernet communication links.

[0008] Furthermore, the dual Ethernet communication links adopt a dual-link isolation mechanism, including several control links and communication links. The communication protocol of each link introduces a link status monitoring and handshake mechanism. Each link is configured with logical role definition, bandwidth limit setting, and priority classification content adjustment through either a JSON file or an XML interface to adapt to different satellite platforms and health management architectures. The data link is used to receive high-frequency telemetry data and output the processing results, anomaly information, and output content of the onboard health management unit; the control link is used to receive control commands and log information.

[0009] Furthermore, the communication link adopts a high-bandwidth TCP / IP long connection and is configured with a zero-copy buffering mechanism to support continuous injection of high-frequency traffic; the control link adopts a high-priority event-driven mechanism, which immediately interrupts the current non-critical task execution process once an instruction packet is detected, and the scheduling processing thread prioritizes parsing and executing the control command.

[0010] Furthermore, the onboard health management unit includes: network port 1, network port 2, write buffer A, buffer switching module, read buffer B, and health management module, which are connected in sequence for communication. Network port 1 is used to receive high-frequency telemetry data frames transmitted via dual Ethernet communication links; Network port 2 is used to receive control commands transmitted via the dual Ethernet communication link; Write buffer A and read buffer B serve as shared memory areas in a heterogeneous computing platform. During the initialization phase of the heterogeneous computing platform, write buffer A is used to receive data frames that have undergone protocol parsing and error correction, and loads the received data frames into read buffer B. When write buffer A becomes full or any of the trigger conditions occur, the buffer switching module marks write buffer A as read and read buffer B as write, and repeats the above mechanism. The health management module is used to bind fixed control links to the corresponding tasks of high-frequency telemetry data, satellite health management unit output, control commands and log information. Tasks are prioritized and scheduled according to the idle status of the links, and the priority of the tasks is determined by a dual-link priority scheduling mechanism for execution.

[0011] A data interaction scheduling method for heterogeneous computing platforms for onboard health management includes: Step 1: At the ground monitoring station, the high-frequency telemetry data frames and control commands are encapsulated and transmitted to the dual Ethernet communication link. The corresponding link in the dual Ethernet communication link introduces a sequence number control and frame loss detection mechanism to the received high-frequency telemetry data frames and control commands and performs verification. Step 2: After verification, the high-frequency telemetry data frame is transmitted to network port 1 through the communication link, and the control command is transmitted to network port 2 through the control link; Step 3: Transmit high-frequency telemetry data frames to the shared memory area through network port 1, write and read data through a dual buffering mechanism, and automatically correct and detect alarms for the stored data through an ECC error correction coding mechanism; Step 4: Bind the corresponding task to the control link through the dual-link priority scheduling mechanism. The task is scheduled according to the priority of the link based on its idle status, and the priority of the task is determined by the dual-link priority scheduling mechanism for sorting and execution.

[0012] Furthermore, step 1 specifically includes: The high-frequency telemetry data frames and control commands of the ground monitoring station adopt a standardized multi-field frame structure. The frame header of the frame structure includes basic information, including but not limited to frame length, frame type, channel ID and sequence number. A CRC-32 field is added to the frame tail for fast error detection. An independent sequence number counter is assigned to each type of high-frequency telemetry data frame, and a 16-32 bit sequence number field is added to the header of each high-frequency telemetry data frame to complete the data encapsulation. The dual Ethernet communication links determine whether an abnormal event has occurred by checking the continuity of sequence numbers. Abnormal events include, but are not limited to, out-of-order frames, lost frames, and duplicate frames. If an abnormal event is detected, it is reported to the heterogeneous computing platform, triggering one of the corresponding fault tolerance and recovery processes. If no abnormal event is detected, a CRC check is performed on each frame of high-frequency telemetry data. After the CRC check is completed, an ACK confirmation is sent to the ground monitoring station. If either the CRC check fails or the frame structure is illegal, the dual Ethernet communication link immediately sends a negative confirmation request. The ground monitoring station re-encapsulates the data and transmits it to the dual Ethernet communication link. If the negative confirmation request does not respond within the timeout period, the ground monitoring station enters the automatic retransmission mechanism, configures the frame granularity and priority of the retransmission request, and sets the preset maximum number of retries and waiting timeout period for the transmission of high-frequency telemetry data frames and control commands. If the current link connecting to the ground monitoring station frequently malfunctions or disconnects, the ground monitoring station will send a switchover request to transfer the data to another redundant link. Once the current link is restored, it will automatically be restored to the main channel and continue to connect with the ground monitoring station.

[0013] Furthermore, the double-buffering mechanism and ECC error correction coding mechanism in step 3 respectively include: During the initialization phase of the heterogeneous computing platform, write buffer A is used to receive data frames that have undergone protocol parsing and error correction, and loads the received data frames into read buffer B. When write buffer A becomes full or any of the trigger conditions occur, the buffer switching module marks read buffer B as a write buffer for writing the protocol-parsed and error-corrected data frames; and marks write buffer A as a read buffer. The ECC error correction coding mechanism automatically corrects single-bit errors during the storage and retrieval phase of data frames and detects and alarms for multi-bit errors.

[0014] Furthermore, step 4 specifically includes: Bind the control command task to the control link and set its priority to P0; bind the onboard health management single-machine output task to the data link and set its priority to P1; bind the high-frequency telemetry data task to the data link and set its priority to P2; bind the log information to the control link and set its priority to P3. Tasks are prioritized based on link availability. Each task is sorted according to its priority before transmission, with higher priority tasks taking over the queue of lower priority tasks. The processing results, anomaly information, and output of the onboard health management unit are then transmitted back to the ground monitoring station via the control link.

[0015] The beneficial effects of this invention are: 1. The "dual-link + priority scheduling" strategy used in this invention greatly shortens the data interaction time. Attached Figure Description

[0016] Figure 1 A structural block diagram of a heterogeneous computing platform for spaceborne health management; Figure 2 A block diagram of the shared memory region; Figure 3 A flowchart illustrating the data interaction scheduling method for a heterogeneous computing platform for onboard health management; Figure 4 This is a flowchart for CRC verification; Figure 5 Timing diagram for single-link serial task processing; Figure 6 This is a timing diagram for handling statically bound links in a dual-link configuration. Figure 7 Timing diagram for dual-link+ optimized scheduling strategy; Figure 8 A schematic diagram of a fault location method based on fault trees; Figure 9A schematic diagram of a lifetime prediction method based on particle filtering and empirical equations; Figure 10 A schematic diagram of a threshold interpretation method based on upper and lower limits; Figure 11 A schematic diagram of an anomaly detection method based on deep learning. Detailed Implementation

[0017] Specific implementation method one: as follows Figure 1 As shown, the structure of the heterogeneous computing platform for onboard health management described in this embodiment includes: The system consists of two interconnected Ethernet communication links and a single onboard health management unit.

[0018] To address the issue of control command congestion caused by data link multiplexing, this implementation employs a dual-link isolation mechanism for the dual Ethernet communication links, including several control links and communication links. Each link's communication protocol incorporates link status monitoring and handshake mechanisms. Each link uses either a JSON file or an XML interface for logical role definition, bandwidth limit setting, and priority classification adjustments to adapt to different satellite platforms and health management architectures. The data link receives high-frequency telemetry data and outputs the processing results, anomaly information, and output content of the onboard health management unit; the control link receives control commands and log information.

[0019] A stable communication link is established based on the TCP / IP protocol to achieve real-time data transmission and interaction between the onboard platform and the ground test / operation control terminal. The dual-channel design effectively avoids congestion on a single link. At the software level, the control link and the communication link are respectively bound to independent network listening services or threads, and each is configured with a parsing protocol stack, data buffering mechanism, and task response process. The data link adopts a high-bandwidth TCP / IP long connection, which has a high tolerance for data access and is configured with a zero-copy buffering mechanism to support continuous injection of high-frequency traffic; while the control link adopts a high-priority event-driven mechanism. Once an instruction packet is detected, the current non-critical task execution process is immediately interrupted, and the scheduling processing thread prioritizes parsing and executing control commands, ensuring that the control channel always has the link guarantee capability of being responsive, interruptible, and capable of backhauling under any data load conditions.

[0020] Furthermore, the link function mapping supports policy file configuration, allowing adjustments to logical role definitions, bandwidth limits, priority classifications, and other aspects via system configuration JSON files or XML interfaces. This enables flexible deployment and customized development across different satellite platforms or health management architectures. To further enhance the operational stability and anomaly handling capabilities of the dual-link mechanism, this invention introduces a link status monitoring and handshake mechanism into the communication protocol design. During each communication cycle, the control link periodically initiates a link health check signal, determining the link's activity status through heartbeat packets, sequence number increment verification, and timeout detection. Once a link disconnection, timeout, or continuous packet anomalies are detected (such as CRC check failures exceeding a threshold), the system automatically switches to degraded mode, suspends high-frequency data injection, and simultaneously transmits error information back to the ground terminal via the control channel, triggering a reconnection process to restore communication. This mechanism ensures that the system retains a certain degree of self-recovery capability and a minimum functional maintenance mechanism even in the event of sudden space radiation disturbances, unexpected link drops, or network congestion.

[0021] The spaceborne health management unit comprises, in sequence, network ports 1 and 2, a write buffer A, a buffer switching module, a read buffer B, and a health management module. At the hardware level, the spaceborne health management unit uses dual gigabit Ethernet ports. Network port 1 receives a large number of high-frequency telemetry data frames injected from the ground or simulation system. Network port 2 receives model control commands, operation status queries and feedback, and parameter configuration modifications from the ground monitoring station, and is responsible for the rapid feedback of processing results, anomaly information, and model output. Figure 2 As shown, the write buffer Buffer A and the read buffer Buffer B serve as shared memory areas of the heterogeneous computing platform. During the initialization phase of the heterogeneous computing platform, the write buffer Buffer A is used to receive data frames that have undergone protocol parsing and error correction, and loads the received data frames into the read buffer Buffer B. When the write buffer Buffer A is full or any of the trigger conditions, the buffer switching module marks the write buffer Buffer A as read and the read buffer Buffer B as write, and repeats the above mechanism.

[0022] The health management module is used to bind fixed control links to the corresponding tasks of high-frequency telemetry data, satellite health management unit output, control commands and log information. Tasks are prioritized and scheduled according to the idle status of the links, and the priority of the tasks is determined by a dual-link priority scheduling mechanism for execution. Specific Implementation Method Two like Figure 3 As shown in this embodiment, the data interaction scheduling method for heterogeneous computing platforms for onboard health management includes: S1: Encapsulate high-frequency telemetry data frames and control commands, and introduce a verification mechanism in the dual Ethernet communication link; To improve the communication reliability of the onboard health management computing platform in a high-interference environment, this invention proposes an integrated anti-interference transmission and verification mechanism. In response to the damage to the stability of the communication link caused by factors such as high-energy particle bombardment, cosmic ray interference, and electromagnetic radiation that spacecraft may encounter during on-orbit operation, multiple protection and recovery strategies are designed to ensure the integrity and recoverability of telemetry data and control commands.

[0024] Specifically, this invention designs a multi-level frame structure and error control. Telemetry and command data are encapsulated during the process, such as... Figure 4 As shown, this implementation adopts a standardized multi-field frame structure. The frame header contains basic information such as frame length, frame type, channel ID, and sequence number, and a CRC-32 (Cyclic Redundancy Check) field is added to the frame tail for fast error detection. CRC check can achieve low-overhead, high-sensitivity error detection of the entire frame content, effectively identifying flip or abrupt errors within 1-4 bits.

[0025] A sequence number control and frame loss detection mechanism is introduced. An independent sequence number counter is assigned to each type of data stream (e.g., telemetry frames, command packets, model output frames), and a 16-32 bit sequence number field is added to the header of each frame. The receiving side uses sequence number continuity to determine whether abnormal events such as out-of-order frames, frame loss, or duplicate frames have occurred. If a sequence number jump or rollback is detected, it is marked as a link anomaly event, which is reported to the system scheduler by the link status monitoring module, triggering corresponding fault tolerance or remediation procedures.

[0026] A lightweight acknowledgment (ACK) protocol is used. After each data frame is successfully received and verified, the receiving end sends an ACK to the sending end. If the CRC check fails or the frame structure is invalid, a NACK (Negative Acknowledgment) request is immediately sent for retransmission. If no response is received within a timeout period, an Automatic Retransmission Request (ARQ) mechanism is initiated, supporting preset maximum retries and timeout periods to prevent prolonged system congestion. In ground testing or simulation environments, the frame granularity (by packet or segment) and priority of retransmission requests can be configured, balancing anti-interference capabilities with latency sensitivity.

[0027] Finally, this implementation method reserves a backup communication channel. In the event of frequent errors or disconnections in the main channel, a switchover request is issued through the link status monitoring module, and transmission is continued via the redundant link. After the link is restored, the system can automatically or manually restore to the main channel, ensuring the continuous availability of the communication link and uninterrupted task operation.

[0028] S2: Upload data and commands via communication and control links; S3: In the shared memory area, data is written and read through a dual-caching mechanism, and the stored data is automatically corrected and detected and alarmed through an ECC error correction coding mechanism; To address the rate mismatch between data reception and model processing during high-speed telemetry injection, a ping-pang dual-buffering mechanism for spaceborne health management platforms is proposed. The core idea of ​​this mechanism is to use two mirror-image buffer regions (Buffer A and Buffer B) alternately, thereby achieving concurrent decoupling of data injection and the health management model without blocking the telemetry reception link.

[0029] Specifically, a shared memory area, named Buffer A and Buffer B, is designed between the telemetry data receiving module and the model calculation module. During system initialization, the current write buffer is set as Buffer A, and the read buffer is set as Buffer B. The telemetry receiving thread continuously writes data frames that have undergone protocol parsing and error correction into Buffer A. When the buffer is full or a condition is triggered, the system immediately swaps the read and write pointers, marking Buffer A as read-only for the model calculation thread to perform data reading and inference calculations; at the same time, Buffer B is cleared and marked as "writable" for receiving subsequent data frames.

[0030] The high-throughput buffer mechanism has the following characteristics: (1) Concurrent execution of data reception and computation: The Ping-Pong structure enables the telemetry receiving thread and the model computation thread to access the same memory area without mutual exclusion, which greatly improves the concurrency and throughput of the processing and avoids data frame loss and lock waiting problems. (2) Frame integrity guarantee and beat alignment: Each buffer rotation is triggered by a fixed frame time window to ensure that the model reads data with consistent length and uniform format, which facilitates batch processing and sliding window mechanism implementation. (3) State management mechanism: The onboard health management unit controls thread synchronization by setting buffer status flags (writable / read-only / idle) to avoid writing overwriting the data being read, and supports error handling logic. If a buffer is detected to have not been switched within a specified time, an abnormal interrupt is triggered to prevent computation blocking.

[0031] In summary, by introducing the aforementioned dual-buffering mechanism, this invention ensures that telemetry data can still be fully received and accurately used by the model even under conditions of high-frequency, high-bandwidth continuous injection, significantly enhancing the system's real-time performance and stability. This is an important foundational structure supporting high-concurrency data stream processing for onboard health management.

[0032] For modules or cache areas in storage, this invention introduces an ECC (Error Correction Code) mechanism to automatically correct single-bit errors and detect and alarm for multi-bit errors during the storage and reading phase, thereby enhancing the stability of data in internal registers or DDR / BRAM.

[0033] S4: Execute tasks by link binding and priority sorting based on the health management module; In spaceborne health management missions, communication objects include not only telemetry data but also various data types such as model control commands, algorithm results, system status feedback, log uploads, and debugging information. These data often share limited communication channel resources, and without a communication scheduling mechanism, problems such as task conflicts, uneven latency, bandwidth saturation, and command blocking can occur. Therefore, this invention provides a health management module deployed on a single machine that includes a fault tree-based fault location method, a particle filter and empirical equation-based lifetime prediction method, a threshold interpretation method based on upper and lower limits, and a deep learning-based anomaly detection method. The flowcharts of these methods are as follows: Figure 8-11 As shown, it is used to quickly interpret, evaluate, and transmit the results of the real-time data stream received from a single machine, thereby enabling fine-grained allocation and dynamic scheduling control of communication tasks.

[0034] The communication scheduling strategy table is shown in Table 1 below.

[0035] Table 1

[0036] Each type of task is pre-bound to a fixed link. For example, telemetry data is bound to a data link for continuous injection of large volumes of telemetry frames; control commands are bound to a control link, which has the highest priority and is used to trigger task start / stop, status query, and parameter configuration; model output is bound to a data link, which has a medium priority and ensures real-time feedback of results; and log information is bound to a control link and is transmitted in a queue as a low-priority task.

[0037] After pre-binding links, tasks are prioritized based on the link's availability. Within each link, the system further introduces a priority-based task sorting mechanism. Before entering the transmission buffer, each type of task is sorted according to its priority level set in the policy table. High-priority tasks (such as control commands) will preempt low-priority tasks (such as logs) from taking their place in the queue, ensuring that critical tasks have an interruptible and responsive transmission channel under any link load conditions.

[0038] After completing the data interaction scheduling using dual-link + priority scheduling, this implementation method compares and tests single-channel serial communication, dual-channel static binding communication, and the dual-link + priority scheduling mechanism to test the "dual-link + priority scheduling" mechanism. The test results are as follows: Figure 5-7 As shown, in traditional communication structures, all task types share a single Ethernet link, tasks are executed serially in chronological order, preemption is not supported, and the processing time exceeds 60 frames. This can lead to uncontrollable control command response time, severe scheduling lag, and delayed model closed-loop output, resulting in a loss of timeliness. In the dual-link scenario, telemetry inputs and outputs are bound to the data link, while control commands and log outputs are bound to the control link. In this case, control commands in the control link may be delayed due to queuing of log commands, model output lags far behind telemetry injection, and the processing time exceeds 50 frames. After introducing the dual-link + priority scheduling method, the shortest path response for control commands is ensured, while tasks that need to be processed are flexibly allocated to idle links, resulting in a processing time of only 37 frames. By comparison, it can be seen that the "dual-link + priority scheduling" strategy used in this invention greatly shortens the data interaction time.

[0039] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent substitutions, and improvements made to the above embodiments without departing from the scope of the present invention, based on the technical essence of the present invention, and within the spirit and principles of the present invention, shall still fall within the protection scope of the present invention.

Claims

1. A heterogeneous computing platform for spaceborne health management, characterized in that, include: Dual Ethernet communication links are used to receive high-frequency telemetry data frames from the ground monitoring station and transmit them to the onboard health management unit, receive control commands issued by the ground monitoring station and transmit them to the onboard health management unit, and transmit the processing results, abnormal information and output content of the onboard health management unit back to the ground monitoring station. The onboard health management unit is bidirectionally connected to the dual Ethernet communication links. The onboard health management unit adopts a dual buffering mechanism, an anti-interference transmission and verification mechanism, and a dual link priority scheduling mechanism to perform the shortest path response for the received control class execution and adaptively allocate the tasks to be processed to the idle links in the dual Ethernet communication links.

2. The heterogeneous computing platform for spaceborne health management according to claim 1, characterized in that, The dual Ethernet communication links adopt a dual-link isolation mechanism, including several control links and communication links. The communication protocol of each link introduces a link status monitoring and handshake mechanism. Each link is configured with either a JSON file or an XML interface to define logical roles, set bandwidth limits, and adjust priority classifications to adapt to different satellite platforms and health management architectures. The data link is used to receive high-frequency telemetry data and output the processing results, anomaly information, and output content of the onboard health management unit. The control link is used to receive control commands and log information.

3. The heterogeneous computing platform for spaceborne health management according to claim 1, characterized in that, The communication link adopts a high-bandwidth TCP / IP long connection and is configured with a zero-copy buffer mechanism to support continuous injection of high-frequency traffic; the control link adopts a high-priority event-driven mechanism. Once an instruction packet is detected, the current non-critical task execution process is immediately interrupted, and the scheduling processing thread prioritizes parsing and executing the control command.

4. The heterogeneous computing platform for spaceborne health management according to claim 1, characterized in that, The onboard health management unit includes: network port 1, network port 2, write buffer A, buffer switching module, read buffer B and health management module, which are connected in sequence for communication. The network port 1 is used to receive high-frequency telemetry data frames transmitted via dual Ethernet communication links. The network port 2 is used to receive control commands transmitted via the dual Ethernet communication link. The write buffer A and read buffer B serve as shared memory areas of the heterogeneous computing platform. During the initialization phase of the heterogeneous computing platform, the write buffer A is used to receive data frames that have undergone protocol parsing and error correction, and load the received data frames into the read buffer B. When the write buffer A is full or any of the trigger conditions, the buffer switching module marks the write buffer A as read and the read buffer B as write, and repeats the above mechanism. The health management module is used to bind fixed control links to the corresponding tasks of high-frequency telemetry data, satellite health management unit output, control commands and log information. Tasks are prioritized and scheduled according to the idle status of the links, and the priority of the tasks is determined by a dual-link priority scheduling mechanism for sorting and execution.

5. A data interaction scheduling method for a heterogeneous computing platform for spaceborne health management, applied to the heterogeneous computing platform for spaceborne health management as described in any one of claims 1-4, characterized in that, include: Step 1: At the ground monitoring station, the high-frequency telemetry data frames and control commands are encapsulated and transmitted to the dual Ethernet communication link. The corresponding link in the dual Ethernet communication link introduces a sequence number control and frame loss detection mechanism to the received high-frequency telemetry data frames and control commands and performs verification. Step 2: After verification, the high-frequency telemetry data frame is transmitted to network port 1 through the communication link, and the control command is transmitted to network port 2 through the control link; Step 3: Transmit high-frequency telemetry data frames to the shared memory area through network port 1, write and read data through a dual buffering mechanism, and automatically correct and detect alarms for the stored data through an ECC error correction coding mechanism; Step 4: Based on the health management module, the corresponding tasks are bound to the control link through the dual-link priority scheduling mechanism. The tasks are scheduled according to the priority of the link based on its idle status, and the priority of the tasks is determined by the dual-link priority scheduling mechanism for sorting and execution.

6. The data interaction scheduling method for heterogeneous computing platforms for onboard health management according to claim 5, characterized in that, Step 1 specifically includes: The high-frequency telemetry data frames and control commands of the ground monitoring station adopt a standardized multi-field frame structure. The frame header of the frame structure includes basic information, including but not limited to frame length, frame type, channel ID and sequence number. A CRC-32 field is added to the frame tail for fast error detection. An independent sequence number counter is assigned to each type of high-frequency telemetry data frame, and a 16-32 bit sequence number field is added to the header of each high-frequency telemetry data frame to complete the data encapsulation. The dual Ethernet communication links determine whether an abnormal event has occurred by checking the continuity of sequence numbers. Abnormal events include, but are not limited to, out-of-order frames, lost frames, and duplicate frames. If an abnormal event is detected, it is reported to the heterogeneous computing platform, triggering one of the corresponding fault tolerance and recovery processes. If no abnormal event is detected, a CRC check is performed on each frame of high-frequency telemetry data. After the CRC check is completed, an ACK confirmation is sent to the ground monitoring station. If either the CRC check fails or the frame structure is illegal, the dual Ethernet communication link immediately sends a negative confirmation request. The ground monitoring station re-encapsulates the data and transmits it to the dual Ethernet communication link. If the negative confirmation request does not respond within the timeout period, the ground monitoring station enters the automatic retransmission mechanism, configures the frame granularity and priority of the retransmission request, and sets the preset maximum number of retries and waiting timeout period for the transmission of high-frequency telemetry data frames and control commands. If the current link connecting to the ground monitoring station frequently malfunctions or disconnects, the ground monitoring station will send a switchover request to transfer the data to another redundant link. Once the current link is restored, it will automatically be restored to the main channel and continue to connect with the ground monitoring station.

7. The data interaction scheduling method for heterogeneous computing platforms for onboard health management according to claim 5, characterized in that, The double buffering mechanism and ECC error correction coding mechanism in step 3 include: During the initialization phase of the heterogeneous computing platform, write buffer A is used to receive data frames that have undergone protocol parsing and error correction, and loads the received data frames into read buffer B. When write buffer A becomes full or any of the trigger conditions occur, the buffer switching module marks read buffer B as a write buffer for writing the protocol-parsed and error-corrected data frames; and marks write buffer A as a read buffer. The ECC error correction coding mechanism automatically corrects single-bit errors during the storage and retrieval phase of data frames and detects and alarms for multi-bit errors.

8. The data interaction scheduling method for heterogeneous computing platforms for onboard health management according to claim 5, characterized in that, Step 4 specifically includes: Bind the control command task to the control link and set its priority to P0; bind the onboard health management single-machine output task to the data link and set its priority to P1; bind the high-frequency telemetry data task to the data link and set its priority to P2; bind the log information to the control link and set its priority to P3. Tasks are prioritized based on link availability. Each task is sorted according to its priority before transmission, with higher priority tasks taking over the queue of lower priority tasks. The processing results, anomaly information, and output of the onboard health management unit are then transmitted back to the ground monitoring station via the control link.

Images