CN-121984879-A - AI-driven heterogeneous cooperative system oriented to network fusion and data processing method

CN121984879ACN 121984879 ACN121984879 ACN 121984879ACN-121984879-A

Abstract

The invention discloses an AI-driven heterogeneous cooperative system oriented to network fusion and a data processing method, comprising the following steps: the system comprises a routing control board card, an AI computing board card, a security board card, an internal and external network synchronous protection module, a software system and a high-speed interconnection bus. The intelligent, safe and highly reliable control of the data transmission process is realized through the deep fusion of hardware and software. The system can complete intelligent link management, dynamic encryption protection, service priority scheduling and real-time data synchronization through multi-calculation-force cooperation and multi-network fusion in a diversified environment so as to support high-safety industry communication scenes of future railways/low altitudes/satellites and the like.

Inventors

ZHENG TAO
LING ZHIMIN
LU BINJIE
ZHANG JIA
LIU BOWEN
HUANG YITIAN
YIN HUAN

Assignees

北京交通大学

Dates

Publication Date: 20260505
Application Date: 20260108

Claims (10)

1. The utility model provides a heterogeneous cooperative system of AI drive towards network integration, system can be simultaneously access two at least in public network, private network, satellite communication network, the short-range communication network, realizes the transmission of many network developments integration based on the decision of AI calculation integrated circuit, its characterized in that includes: the routing control board is internally provided with a multi-core heterogeneous processor and is used for executing network protocol analysis, data forwarding, multi-network link control and QoS scheduling; The AI computing board card integrates at least one computing core in GPU, NPU, DPU, runs a service classification model, a flow prediction model, a heterogeneous network cooperative model and a safety compliance examination model and is used for generating network resource allocation and transmission path decision and data compliance examination results; the security board card is used for realizing hardware-level encryption, decryption, identity authentication and key updating, constructing an end-to-end trusted transmission link and carrying out security verification on data entering and exiting an intranet; The internal and external network synchronous protection module is arranged between an internal network and an external network, the internal network is a security domain for bearing a service data stream needing to be protected, and the external network is a fusion transmission domain comprising at least two of a public network, a private network, a satellite communication network and a near-distance communication network; the software system adopts a multi-operating system cascading structure, different boards run independent OSs, and cross-system communication, task synchronization and data consistency maintenance are realized through a middleware layer; The routing control board card, the AI computing board card and the security board card are connected through the high-speed interconnection bus, and realize high-speed data exchange and synchronization through the shared memory area and the cross-domain data buffer mechanism.
2. The network fusion-oriented AI-driven heterogeneous collaborative system according to claim 1, wherein a multi-core heterogeneous processor built in a routing control board is a multi-core heterogeneous architecture, which runs an operating system and supports differentiated scheduling of heterogeneous multi-source data; The AI computing board card comprises a multi-model fusion algorithm, a decision weight, a video feedback and entertainment data, a video feedback module and a video feedback module, wherein the decision weight can be dynamically adjusted according to the service type, the security service is higher in priority than the non-security service, and the security service comprises a control signal and a scheduling instruction; the security board card supports at least one encryption algorithm in AES, RSA, SM and SM9, and the key updating period can be dynamically adjusted based on the network state; the high speed interconnect bus includes a PCIe, serDes, or CCIX bus.
3. The network fusion-oriented AI-driven heterogeneous cooperative system of claim 1, wherein the inner network and the outer network are synchronously protected by a protection module for realizing data isolation and mapping of the inner network and the outer network, and preventing information leakage, wherein the inner network comprises a security domain, and the outer network comprises a transmission domain.
4. The AI-driven heterogeneous collaboration system of claim 1, wherein the near-field communication network comprises at least one of Wi-Fi, loRa, bluetooth, and unmanned self-organizing network.
5. The AI-driven heterogeneous collaboration system for network fusion of claim 1, wherein the system is adapted for use in at least one of a vehicular communication network, an earth-of-rail data transmission system, a low-altitude intelligent networking data transmission system, and a satellite communication network.
6. The AI-driven heterogeneous collaboration system of claim 1, further comprising an intranet-extranet synchronization protection mechanism, wherein the mechanism constructs a collaboration protection system of a security domain and a transmission domain based on a hardware architecture, and specifically comprises: The intranet is a security domain and is used for bearing a service data stream needing security protection, and the stream of the service data stream is sequentially subjected to encryption processing of a security board card and security compliance model examination of an AI computing board card, and is transmitted to a routing control board card after the examination is passed; The external network is a transmission domain, and at least two of a public network, a private network, a satellite communication network and a near communication network form a converged network for receiving the encrypted service data stream forwarded by the routing control board card; The internal and external network synchronous protection mechanism further comprises an internal and external network synchronous protection module which is deployed between the internal network and the external network and is used for: 1) The encryption strategy of the internal network security board card and the security state of the external network transmission link are synchronized in real time, so that the encryption algorithm, the key version and the protection requirement of the external network link are matched; 2) The feedback data returned from the external network to the internal network is reversely checked, namely, the data is decrypted through the security board card, the data integrity and the source legitimacy are checked through the AI calculation board card, and the data is allowed to enter the internal network security domain after the check is passed; 3) Illegal data which does not pass through AI model examination and encrypted data with unmatched secret keys are intercepted, and meanwhile, an abnormal log is generated and an alarm is triggered, so that the whole process of the internal and external network data interaction is controllable and traceable.
7. The network convergence oriented AI-driven heterogeneous collaboration system of claim 6, wherein the security compliance model inspection of the AI computing board comprises inspection of service type validity, data format integrity, transmission priority matching degree of the service data flow, and the data which is not passed through inspection is returned to an intranet source and is not allowed to enter a route forwarding flow, wherein the service type validity inspection of the service data flow comprises inspection of whether the service type validity inspection belongs to a preset security service list, the data format integrity inspection comprises inspection of whether the service type validity inspection accords with encryption transmission specifications, and the transmission priority matching degree inspection comprises inspection whether the service type validity inspection is matched with an extranet link bandwidth and a time delay index.
8. The AI-driven heterogeneous collaborative data processing method for network fusion is characterized by comprising the following steps: The method comprises the steps that S1, business data are encrypted and checked in integrity through a security board card, wherein the step corresponds to a security board card processing link, namely, the business data firstly enter the security board card, encryption and checked in integrity are performed at the security board card, and original security of the data is ensured; The step 2 is that the encrypted data is transmitted to an AI computing board card, and the AI model in the AI computing board card analyzes the network state, the service priority and the bandwidth information to generate an optimal transmission strategy, wherein the step corresponds to an AI computing board card decision link, namely the encrypted data is transmitted to the AI computing board card later, and before routing, the AI model comprehensively analyzes the network state, the service priority and the bandwidth information to dynamically generate the optimal transmission strategy instead of the traditional static routing configuration; The step 3 is that the routing control board card executes multi-network scheduling and load balancing according to the transmission strategy to realize data fusion transmission, and the step corresponds to the execution link of the routing control board card, namely the routing control board card receives encrypted data and executes specific multi-network scheduling and load balancing operation according to the transmission strategy generated by the AI calculation board card strictly to realize fusion transmission of the data in heterogeneous networks; And S4, the receiving end performs reverse decryption and verification through the secure board card to finish data distribution, wherein the step corresponds to a processing link of the receiving end, namely, after the data reaches the receiving end through heterogeneous network transmission, the receiving end performs reverse decryption and verification through the secure board card of the receiving end to finish data secure distribution finally.
9. The network fusion-oriented AI-driven heterogeneous cooperative data processing method of claim 8, wherein the network fusion-oriented AI-driven heterogeneous cooperative intelligent router system architecture comprises sequential logic under three scenarios, namely normal data transmission, abnormal blocking and dynamic policy adjustment; The first scenario is that data is unidirectionally transmitted, the data must be forced to pass through FPGA, GPU, CPU in sequence, if any link is disconnected or refused, the data cannot reach the external network, and the method comprises the following steps of sending original sensitive data, conveying encrypted data flow, passing examination, then flowing to a routing layer, and integrating and distributing to the external network; the second scene is that the fusing function of AI cuts off the path flowing to CPU and sends alarm to the management plane when GPU detects abnormality, which comprises intercepting interrupt transmission; The third scenario is that the system is intelligent, when the CPU senses that the external network environment is poor, the system does not wait passively, but actively triggers the Sync agent, the Sync agent then instructs the three boards in parallel to cooperatively adjust, the FPGA reduces the encryption load to exchange speed, the AI adjusts the inspection strategy, the CPU switches the physical link, and the method comprises the steps of reporting network state data, switching to a low bandwidth overhead encryption algorithm, adjusting the inspection threshold to keep alive preferentially, and updating the routing weight to switch to the satellite link.
10. The AI-driven heterogeneous collaborative data processing method for network fusion according to claim 8, wherein the AI model in step S2 predicts network traffic state in real time, adjusts transmission path in advance, reduces data packet loss rate, and the multi-network scheduling in step S3 supports link failure fast switching and has link-level redundancy backup capability.

Description

AI-driven heterogeneous cooperative system oriented to network fusion and data processing method Technical Field The invention relates to the technical field of communication, in particular to an AI-driven heterogeneous cooperative system oriented to network fusion and a data processing method. Background With the rapid evolution of the intelligent networking automobile, the track traffic (including heavy load transportation), the low-altitude intelligent networking (such as unmanned aerial vehicle cluster), the satellite communication and other fields towards the intelligent, unmanned and high-reliability directions, the data transmission requirements under each scene are characterized by double-surge, on one hand, the data interaction quantity of the vehicle-mounted environment (vehicle-road cooperative data), the track traffic vehicle ground (real-time scheduling/state monitoring/video backhaul of heavy load trains), the low-altitude intelligent networking (unmanned aerial vehicle cluster control/high-definition data backhaul) and the satellite communication (space-ground integrated data interaction) is exponentially increased, and on the other hand, the requirements of the low-delay stability, the high-reliability redundancy and the safety protection level (such as data encryption, electromagnetic interference resistance and link hijack resistance) of the data transmission are remarkably improved, so that the data interaction quantity is a core bottleneck restricting the technology landing in each field. The related communication system and the core routing device of the current mainstream still have the following two key technical defects in two major aspects, and cannot adapt to the requirements: 1. inherent drawbacks of existing network architecture are the difficulty in supporting the "converged transmission" requirement Currently, a single network architecture or a multi-network discrete deployment mode is generally adopted in a vehicle-mounted communication network, a railway LTE-R vehicle-ground system, a low-altitude local communication network, a satellite ground access network and the like, so that three major core problems are caused: (1) The reliability is insufficient, namely, a single network is easily influenced by scenerization interference (such as track traffic tunnel signal attenuation, low-altitude complex electromagnetic environment and vehicle-mounted mobile channel shielding) or link interruption (such as satellite link rain attenuation) and failure redundancy cannot be provided, and when multiple networks are separately deployed, each network independently operates and resources are split, and link state sensing and dynamic resource allocation cannot be realized (such as when a heavy-duty train enters a tunnel, the LTE-R cannot be quickly switched to other backup links after interruption). (2) The delay uncertainty is that a single network architecture is difficult to avoid sudden delay fluctuation in the face of real-time services (such as vehicle-mounted train path cooperative control signals, heavy-load train dispatching instructions and unmanned aerial vehicle real-time control data), and a discrete architecture is more incapable of ensuring delay stability through multi-network cooperation, so that service response lag risks are caused. (3) The existing scheme lacks an integrated safety mechanism adapting to multiple scenes, namely the vehicle-mounted scenes are easy to face the data tampering of the internet of vehicles, the track traffic scenes need to prevent signal transmission hijacking, the low-altitude intelligent networking needs to resist electromagnetic interference, the satellite communication needs to ensure sensitive data encryption, a distributed safety strategy under a discrete architecture cannot form a protection closed loop, and the security holes are prominent. 2. The existing controller has the problems of difficult support of 'cooperative intelligence' requirement due to calculation force and functional bottleneck As core devices (such as a vehicle-mounted router, a vehicle-ground communication controller, a low-altitude intelligent networking gateway and the like) of each communication system, the existing routing devices generally adopt a centralized control structure of a single CPU, only can complete basic network protocol processing and data forwarding, and three key requirements of 'heterogeneous computation force coordination, AI intelligent decision and hardware-level security' cannot be considered, and the specific bottlenecks are as follows: (1) Heterogeneous computing power synergetic defects are that a single CPU cannot integrate heterogeneous computing power resources of general Computing (CPU), AI computing (GPU/TPU used for route decision optimization) and special encryption computing (hardware encryption chip), so that two problems are caused, namely, the AI computing power is insufficient, an intelligent routing strategy cannot be genera