Aircraft heterogeneous node efficient data distribution middleware

Abstract

The aircraft heterogeneous node efficient data distribution middleware is composed of a two-stage structure, wherein the two-stage structure comprises an in-chip lightweight communication kernel and an inter-chip interface packaging soft bus; the in-chip lightweight communication kernel comprises a kernel driving module, a shared memory, a signal notification mechanism and a user space client library; the inter-chip interface packaging soft bus comprises a unified message packaging device, an interface adapter, a message routing module and a message bridging adapter. The invention aims to solve the problems of communication mechanism splitting, insufficient real-time performance, static solidification of system topology and the like among multi-chip and multi-architecture processors in an aircraft system. The middleware supports data frame registration and unbinding, broadcast scheduling and node adaptive updating during operation, has the characteristics of high throughput, low delay, strong extension, interface independence and the like, and is suitable for typical heterogeneous embedded systems such as aircraft task calculation, flight control navigation, image recognition and the like.

Description

Technical Field

[0001] This invention belongs to the field of communication architecture technology of aerospace electronics, embedded systems and real-time operating systems, and specifically relates to an efficient data distribution middleware for heterogeneous nodes of aircraft. Background Technology

[0002] As the complexity of aircraft platform missions increases, aircraft systems integrate a variety of heterogeneous processing nodes, including main control SoC, signal processing DSP, image recognition accelerator, navigation and measurement module, fault diagnosis unit, etc. These chips usually run on their own independent processing architecture and operating system environment (such as SylixOS, Tianmai OS, bare metal, etc.). The systems need to frequently, at high speed and with low latency exchange mission control information, status feedback, sensor data and execution results.

[0003] Currently, data communication between heterogeneous nodes in aircraft generally suffers from the following problems:

[0004] 1) Fragmented communication mechanisms: The interface types between chips are not uniform (SPI, CAN, serial port, Ethernet, shared bus, etc.), and the software interfaces are different, making it difficult to build a unified communication architecture;

[0005] 2) Low data interaction efficiency: It relies heavily on polling or blocking read mechanisms, resulting in insufficient real-time performance and resource utilization, and messages between tasks cannot be accurately distributed.

[0006] 3) The system topology is statically fixed: the communication path and binding relationship are set at compile time, which makes it difficult to adapt to dynamic adjustment of task mode and "hot-plugging" of subsystems;

[0007] 4) Lack of cross-node awareness: There is no unified addressing and message orchestration system among the devices, which can easily cause disorder in command issuance and feedback collection, increasing the complexity of collaboration.

[0008] Therefore, there is an urgent need for a unified and efficient communication middleware for heterogeneous aircraft systems that can provide a cross-chip and cross-system task data collaboration mechanism without sacrificing on-chip performance. Summary of the Invention

[0009] The purpose of this invention is to provide an efficient data distribution middleware for heterogeneous nodes in aircraft, which solves the problems of fragmented communication mechanisms, insufficient real-time performance, and statically fixed system topology among multiple chips and processors within existing aircraft.

[0010] To achieve the above objectives, this invention provides an efficient data distribution middleware for heterogeneous nodes in aircraft, consisting of a two-level structure: an in-chip lightweight communication kernel and an inter-chip interface encapsulated soft bus. The in-chip lightweight communication kernel includes: a kernel driver module: registering the character device / dev / softbus in each chip's operating system, serving as the bus scheduling center for that node; shared memory: allocating a memory region by the driver when each data frame is created, for shared reading and writing by processes; a signal notification mechanism: data updates are actively notified to consumers by the driver using custom signals, completely eliminating polling and blocking waiting; and a user-space client library: encapsulating all communication operations into a unified user interface (API), transparent to the application layer. The inter-chip interface encapsulation soft bus includes: a unified message encapsulator: encapsulates local data frames into standard message packet format, realizing message serialization, unpacking, and verification functions; an interface adapter: abstracts various physical links, uniformly providing sendToNode() and receiveFromNode() methods; a message routing module: each node holds a dynamic node topology table and a message routing rule table, and searches for the target node from the node topology table based on the message destination address, forwarding it to the corresponding physical interface; and a message bridging adapter: local chip data frames can be registered as "exportable message sources" through the bridging component, driving the encapsulation and forwarding of the data to other nodes when data is updated, completing the outward distribution of messages.

[0011] The aforementioned high-efficiency data distribution middleware for heterogeneous aircraft nodes includes a kernel driver module that acts as a unified message arbitrator, responsible for data frame lifecycle management, access control, and signal event sending. The shared memory utilizes shared memory mapping and cross-process read-write lock mechanisms to achieve zero-copy shared access between processes. The signal notification mechanism sends signals carrying frame index parameters, which consumers use to accurately locate updated data frames. The user-space client library allows the application layer to dynamically register, search, bind, read, write, unregister, and read data frames through a unified interface.

[0012] The aforementioned high-efficiency data distribution middleware for heterogeneous nodes of aircraft includes a lightweight communication kernel within the chip that provides microsecond-level message delivery latency, supports multiple producers and multiple consumers concurrently, and is suitable for embedded RTOS scenarios.

[0013] The aforementioned high-efficiency data distribution middleware for heterogeneous aircraft nodes includes an interface adapter that supports transmission over multiple physical links and shields underlying differences through a unified interface adapter; a message routing module that forwards or broadcasts cross-node messages based on a runtime updatable node topology table; and a message bridging adapter that can bridge local chip data frames into cross-node messages, forward them to heterogeneous nodes, and decapsulate and restore them to local chip data frames at the receiving end.

[0014] The aforementioned high-efficiency data distribution middleware for heterogeneous aircraft nodes includes, in the form of, the inter-chip interface encapsulation soft bus, which further includes: a heterogeneous adaptation mechanism: providing lightweight polling or interrupt-driven alternative access methods for bare metal nodes or nodes that do not support mmap and signal systems.

[0015] The aforementioned high-efficiency data distribution middleware for heterogeneous nodes of aircraft supports broadcast mode; when the target node is unreachable, it falls back to broadcast mode.

[0016] The aforementioned high-efficiency data distribution middleware for heterogeneous aircraft nodes allows keyframes to be set as "global broadcast frames" based on system usage requirements, automatically forwarded to all registered listening nodes.

[0017] The aforementioned high-efficiency data distribution middleware for heterogeneous aircraft nodes employs message sequence numbering and acknowledgment (ACK) mechanisms to prevent erroneous retransmissions and out-of-order delivery; it supports runtime frame addition and deletion, node online / offline detection, and enables real-time updates of the topology graph.

[0018] The aforementioned high-efficiency data distribution middleware for heterogeneous aircraft nodes allows for direct deployment of this middleware on all nodes running SylixOS or other POSIX API-supporting nodes. In rudder control-type small-core devices, a read-only image frame + polling method is used to ensure controlled power consumption and resource overhead. All communication can be statically configured via configuration tables or dynamically registered at runtime, supporting task decoupling and hot-swappable service deployment.

[0019] The aforementioned high-efficiency data distribution middleware for heterogeneous aircraft nodes provides a trimmed version of SoftBus Client for bare-metal or RTOS nodes, using interrupts or DMA to replace signals and mmap.

[0020] Compared with the prior art, the beneficial technical effects of the present invention are:

[0021] 1) Unified bus abstract model: Cross-node communication is based on the "message frame" model, providing a unified view for communication within and across chips, thus improving the overall system maintainability;

[0022] 2) Low latency and high throughput: In-chip communication can reach millions of message QPS, cross-chip communication has high packaging efficiency, and it is suitable for real-time task control and data synchronization requirements;

[0023] 3) Dynamic topology adaptation: The system supports adding / unregistering frames and nodes at runtime to meet the needs of dynamic mode switching, function pruning and fault reconstruction scenarios;

[0024] 4) Hardware interface independence: The underlying communication link is abstracted, which can be adapted to different hardware interfaces and supports rapid access to heterogeneous SoC platforms.

[0025] 5) Cross-scenario general capability: In addition to aircraft, the middleware of this invention is also applicable to typical heterogeneous embedded platforms such as edge computing devices, UAV systems, and airborne radar subsystems. Attached Figure Description

[0026] The efficient data distribution middleware for heterogeneous nodes of aircraft of the present invention is given by the following embodiments and figures.

[0027] Fig. 1 This is a typical logical structure diagram of the efficient data distribution middleware for heterogeneous nodes of an aircraft, as described in an embodiment of the present invention.

[0028] Fig. 2 This is a schematic diagram of the cross-node message encapsulation format in an embodiment of the present invention.

[0029] Fig. 3 This is a flowchart of the dynamic message routing process in an embodiment of the present invention. Detailed Implementation

[0030] The following will combine Figs. 1-3 The efficient data distribution middleware for heterogeneous nodes of aircraft of the present invention will be described in further detail.

[0031] like Fig. 1 The efficient data distribution middleware for heterogeneous nodes of aircraft designed in this embodiment of the invention consists of a two-level structure:

[0032] A) On-chip lightweight communication kernel (SoftBus) – the core mechanism of on-chip communication

[0033] Within each node (chip), a lightweight on-chip communication kernel is used as the underlying communication engine, specifically including:

[0034] Kernel driver module: Registers the character device / dev / softbus in each chip's operating system, serving as the bus scheduling center for that node; the kernel driver module acts as a unified message arbitrator, responsible for data frame lifecycle management, access control, signal event sending, etc.

[0035] Shared memory: When each data frame is created, the driver allocates a memory region for processes to read and write; zero-copy shared access between processes is achieved using shared memory mapping (mmap) and cross-process read-write lock mechanism (pthread_rwlock_t);

[0036] Signal notification mechanism: Data updates are actively notified to consumers by the driver using a custom signal (sigqueue), completely eliminating polling and blocking waiting; the sent signal carries a frame index parameter, which consumers can use to accurately find the updated data frame;

[0037] User space client library: Encapsulates all communication operations into a unified user operation interface (API) for transparent use by the application layer. That is, the application layer can perform operations such as dynamic registration, search, binding, reading, writing, deregistration and reading of data frames through the unified interface;

[0038] The lightweight communication core within the chip can provide microsecond-level message delivery latency and has multi-producer-multi-consumer concurrency support capabilities, making it suitable for embedded real-time operating system (RTOS) scenarios.

[0039] B) Inter-chip interface packaged soft bus – an inter-chip communication extension mechanism

[0040] To support cross-node data synchronization and collaboration, a unified inter-chip interface encapsulation software bus is proposed, specifically including:

[0041] Unified Message Encapsulator: Encapsulates local data frames into standard message packet format (frame header + source address + destination address + frame type + payload data + checksum), and implements message serialization, unpacking and verification functions;

[0042] Interface Adapter: Abstracts physical links such as SPI, CAN, serial port, and Ethernet, and provides unified sendToNode() and receiveFromNode() methods; supports transmission on multiple hardware links, and shields the differences of the underlying layers through a unified interface adapter;

[0043] Message routing module: Each node holds a dynamic node topology table and a message routing rule table. Based on the destination address of the message, it looks up the target node from the node topology table and forwards it to the corresponding physical interface. That is, the message routing module forwards or broadcasts cross-node messages according to the node topology table that can be updated at runtime.

[0044] Message bridging adapter: Local chip data frames can be registered as "exportable message sources" through the bridging component. The driver encapsulates them and forwards them to other nodes when the data is updated, thus completing the outward distribution of messages. That is, the message bridging adapter can bridge local chip data frames into cross-node messages, forward them to heterogeneous nodes, and decapsulate them at the receiving end to restore them to local chip data frames.

[0045] Heterogeneous adaptation mechanism: For bare metal or nodes that do not support mmap and signal systems, a lightweight polling or interrupt-driven alternative access method is provided.

[0046] The middleware designed in this embodiment achieves seamless unification of message paths inside and outside the chip, and builds a lightweight middleware platform that supports task decoupling, data broadcasting, and status publishing.

[0047] The working principle of the middleware designed in this embodiment is as follows:

[0048] 1) Data frame creation and binding

[0049] Taking the creation and binding of data frames within the main control SoC chip as an example, it specifically includes:

[0050] Application process A calls SoftBusClient::createDataFrame() to create a "control command frame" of 64 bytes;

[0051] The kernel driver module allocates a shared memory block, initializes pthread_rwlock_t, and returns the frame index;

[0052] Application process B calls findDataFrame() to locate the frame and calls bindConsumer() to bind it as a consumer;

[0053] The driver maps shared memory and registers the consumer as the target, completing the internal publish-subscribe configuration.

[0054] Process A calls writeDataFrame() to write a command, which triggers a broadcast SIG_DATA_WRITE signal to notify process B.

[0055] The callback function of process B is automatically triggered, reads the command content from shared memory and processes it;

[0056] 2) Encapsulation and forwarding of inter-chip messages

[0057] Taking the sending of target point information from the main control SoC chip to the navigation chip as an example, the specific steps include:

[0058] The main control SoC chip application process A calls writeDataFrame() to write telemetry data to the frame "RemoteCommand";

[0059] The message bridge adapter detects that the frame is an "exportable message source" and triggers the message routing module;

[0060] The unified message encapsulator encapsulates the frame content into a standard message packet format (frame header + source address + destination address + frame type + payload data + checksum, such as...). Fig. 2 );

[0061] The message routing module looks up the table, finds the target node such as the "data and remote control integration module", and hands it over to its physical interface adaptation layer (such as LVDS);

[0062] The LVDS adapter writes the message into the transmission FIFO and sends it to the data and remote sensing module;

[0063] After receiving the data, the middleware framework unpacks it, restores it to a local data frame, and writes it into its SoftBus shared memory.

[0064] The telemetry control program process receives the notification signal of this frame through binding, reads the data, and then executes telemetry transmission;

[0065] 3) Dynamic message routing and broadcasting mechanism

[0066] Each node middleware maintains a dynamic node topology table in a distributed manner, including a local "frame table" and a "neighbor node table";

[0067] If the target node is unreachable, it can fall back to broadcast mode;

[0068] A message sequence number and acknowledgment (ACK) mechanism is used to prevent erroneous retransmissions and out-of-order delivery;

[0069] It supports adding and deleting frames at runtime, and detecting nodes going online or offline, so as to achieve real-time updates of the topology graph;

[0070] Depending on the system's usage requirements, keyframes can be set as "global broadcast frames" and automatically forwarded to all registered listening nodes;

[0071] A typical message passing process is as follows: message to be sent → written to softbus → decide whether to send externally or transfer within the chip based on whether it is a local frame → find the external interface based on the neighbor table and frame table → send externally. See details in [link to process details]. Fig. 3 .

[0072] Middleware deployment and compatibility adaptation in this embodiment:

[0073] This middleware can be deployed directly on any node running SylixOS or other nodes that support the POSIX API;

[0074] For bare-metal / RTOS nodes, a trimmed version of the SoftBus Client is provided, which uses interrupts or DMA to replace signals and mmap.

[0075] In small core devices with rudder control, a read-only image frame + polling reading method is used to ensure that power consumption and resource overhead are controlled.

[0076] All communication can be statically configured through a configuration table or dynamically registered at runtime, supporting decoupling between tasks and hot-swappable service deployment.

[0077] This invention constructs a lightweight communication kernel (SoftBus) within the chip, achieving zero-copy multi-process communication through shared memory and signal notification mechanisms. It provides unified message encapsulation, physical interface abstraction, and dynamic routing mechanisms between chips, enabling efficient data distribution and bridging synchronization across nodes. The middleware supports runtime data frame registration and unbinding, broadcast scheduling, and node adaptive updates, featuring high throughput, low latency, strong scalability, and interface independence. It is suitable for typical heterogeneous embedded systems such as aircraft mission computing, flight control and navigation, and image recognition.

[0078] This invention addresses the data interaction needs between the main control chip and various coprocessor chips and devices. It integrates a soft bus mechanism, unified message addressing, cross-chip communication interface encapsulation, and dynamic message routing mechanism to achieve efficient, low-latency, scalable, and hot-swappable data publishing and subscription capabilities between heterogeneous nodes.

[0079] The efficient data distribution middleware for heterogeneous nodes in aircraft of the present invention is suitable for communication of heterogeneous systems in highly dynamic and high-density constrained environments, such as communication between multiple chips and processors with multiple architectures inside an aircraft.

Claims

1. A high-efficiency data distribution middleware for heterogeneous nodes in aircraft, characterized in that: It consists of a two-level structure: an on-chip lightweight communication core and an inter-chip interface packaged soft bus; The lightweight communication core within the chip includes: Kernel driver module: Registers the character device / dev / softbus in each chip's operating system, serving as the bus scheduling center for that node; Shared memory: When each data frame is created, the driver allocates a memory region for processes to read and write. Signal notification mechanism: Data updates are proactively notified to consumers by the driver using custom signals, completely eliminating polling and blocking waiting; User space client library: Encapsulates all communication operations into a unified user operation interface API, which can be used transparently by the application layer; The inter-chip interface packaged soft bus includes: Unified Message Encapsulator: Encapsulates local data frames into standard message packet format, enabling message serialization, unpacking, and verification functions; Interface adapter: Abstracts various physical links and provides unified sendToNode() and receiveFromNode() methods; Message routing module: Each node holds a dynamic node topology table and a message routing rule table. Based on the destination address of the message, it looks up the target node from the node topology table and forwards it to the corresponding physical interface. Message bridging adapter: Data frames within the local chip can be registered as "exportable message sources" through the bridging component. When the data is updated, the driver encapsulates the data and forwards it to other nodes, thus completing the outward distribution of messages.

2. The high-efficiency data distribution middleware for heterogeneous nodes of aircraft as described in claim 1, characterized in that, The kernel driver module acts as a unified message arbiter, responsible for data frame lifecycle management, access control, and signal event sending; the shared memory uses shared memory mapping and cross-process read-write lock mechanisms to achieve zero-copy shared access between processes; the signal notification mechanism sends signals carrying frame index parameters, which consumers use to accurately locate updated data frames. The user space client library allows the application layer to dynamically register, search, bind, read, write, unregister, and read data frames through a unified interface.

3. The high-efficiency data distribution middleware for heterogeneous nodes of aircraft as described in claim 1, characterized in that, The lightweight communication core within the chip provides microsecond-level message delivery latency and supports multiple producer-consumer concurrency, making it suitable for embedded real-time operating system (RTOS) scenarios.

4. The high-efficiency data distribution middleware for heterogeneous nodes of aircraft as described in claim 1, characterized in that, The interface adapter supports transmission over multiple physical links, shielding underlying differences through a unified interface adapter; the message routing module forwards or broadcasts cross-node messages based on a runtime updatable node topology table. It can bridge local chip data frames into cross-node messages through a message bridging adapter, forward them to heterogeneous nodes, and decapsulate and restore them to local chip data frames at the receiving end.

5. The high-efficiency data distribution middleware for heterogeneous nodes of aircraft as described in claim 1, characterized in that, The inter-chip interface packaged soft bus also includes: Heterogeneous adaptation mechanism: For bare metal or nodes that do not support mmap and signal systems, a lightweight polling or interrupt-driven alternative access method is provided.

6. The high-efficiency data distribution middleware for heterogeneous nodes of aircraft as described in claim 1, characterized in that, The middleware supports broadcast mode; it falls back to broadcast mode when the target node is unreachable.

7. The high-efficiency data distribution middleware for heterogeneous nodes of aircraft as described in claim 1, characterized in that, Depending on the system's usage requirements, keyframes can be set as "global broadcast frames" and automatically forwarded to all registered listening nodes.

8. The high-efficiency data distribution middleware for heterogeneous nodes of aircraft as described in claim 1, characterized in that, The system employs message sequence numbers and ACK mechanisms to prevent erroneous retransmissions and out-of-order delivery; it supports runtime frame addition and deletion, node online / offline detection, and enables real-time updates of the topology graph.

9. The high-efficiency data distribution middleware for heterogeneous nodes of aircraft as described in claim 1, characterized in that, All nodes running SylixOS or other POSIX API-supporting nodes can directly deploy the middleware; in rudder-controlled small-core devices, read-only image frames + polling reading method are used to ensure that power consumption and resource overhead are controlled; all communication can be statically set through configuration tables or dynamically registered at runtime, supporting decoupling between tasks and hot-swappable service deployment.

10. The high-efficiency data distribution middleware for heterogeneous nodes of aircraft as described in claim 9, characterized in that, For bare-metal or RTOS nodes, a trimmed version of the SoftBus Client is provided, which uses interrupts or DMA to replace signals and mmap.

Images