Self-adaptive network construction method oriented to own intelligence

Abstract

The invention provides a self-adaptive network construction method oriented to intelligence, and belongs to the technical field of artificial intelligence. The method comprises the following steps: predicting a network state by using an artificial intelligence algorithm LSTM (Long Short-Term Memory Network), converting a service intention into an executable network strategy according to an intelligent service demand, dynamically optimizing a network path, and establishing a self-adaptive network according to the network path. And an optimal and deterministic network path is provided for the intelligent robot. According to the method, the wide area network based on software definition is optimized, and transmission guarantee is provided for business requirements of cooperation, clustering and the like of the intelligent terminal under the wide area network.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence, and particularly relates to an adaptive network construction method for embodied intelligence. BACKGROUND

[0002] Due to the large demand for computing power of most current embodied intelligence, the computing brain is mostly deployed on a cloud high-performance server, and is sensitive to data transmission requirements when performing a job. Slight jitter, delay, and network drop can cause embodied intelligence to be stuck, gait imbalance, and gesture imbalance, trigger collision or task implementation failure, and affect user experience and normal operation of the device.

[0003] To solve the above problems, some manufacturers use a 5G and WiFi dual-mode set, a network mutual switching mode, and switch to a relatively stable network in different scenes. This scheme can only manually switch when the network sends a fault, and cannot provide a reliable network in a timely manner. Some manufacturers use a 5G+WiFi fusion scheme to further improve network reliability, but this is a great challenge to the power consumption of embodied intelligence. SUMMARY

[0004] To solve the above technical problems, the present application provides an adaptive network construction method for embodied intelligence. The method solves the harsh demand of embodied intelligence on communication, overcomes the jitter and uncertainty of 5G and WiFi, and predicts the future state of the network when no network exception occurs, provides timely link switching and protection strategies, and ensures the stability and certainty of subsequent business.

[0005] The technical scheme of the present application is as follows:

[0006] An adaptive network construction method for embodied intelligence, which uses an artificial intelligence model to predict the network state, and then converts the business intent into an executable network strategy according to the business requirements of embodied intelligence, dynamically optimizes the network path, and provides the best and deterministic network path for embodied intelligence.

[0007] Further,

[0008] By using the business semantic information (such as task type, instruction priority) and network state information of the robot control center, an artificial intelligence model is used to predict the future network state, and based on the prediction result and business strategy, the optimal network path is dynamically selected for different business flows.

[0009] The business semantic information includes but is not limited to robot task type, data flow priority, and maximum delay requirement.

[0010] The artificial intelligence model uses a time series prediction model (LSTM) to predict the delay, jitter, and packet loss rate of the WAN link.

[0011] Further,

[0012] Comprise the following partial modules:

[0013] Network control service module, management of multiple terminal devices, monitoring network state, implementation of network policy, perception of business;

[0014] Embodied intelligent terminal communication client, with WiFi+5G module, and has the functions of double sending and receiving, link aggregation, can execute the network control strategy issued by the embodied intelligent control center;

[0015] Embodied intelligent control center, responsible for all embodied intelligent task scheduling, state monitoring and business logic.

[0016] Among them,

[0017] Network control service module:

[0018] Gathers all the network and business state data reported by the terminal, constructs the global real-time network view and business view; uses machine learning time series prediction model, inputs historical and real-time link performance data, time, network load; predicts the performance index probability distribution of each available WAN link in the future time window;

[0019] Receive the business intent from the embodied intelligent control center;

[0020] Integrate AI prediction results, real-time network state and business strategy, make multi-objective optimization decision, calculate the global optimal traffic path scheme at the current time;

[0021] Real-time and reliable path strategy generated by decision-making is issued to the embodied intelligent terminal communication device client.

[0022] Embodied intelligent terminal communication client:

[0023] Network state acquisition: real-time monitoring and reporting of network indicators in the local embodied intelligence, including: end-to-end delay, jitter, packet loss rate and the state of available network interfaces;

[0024] Receive and strictly execute the path strategy from the network controller, send through the specified network interface.

[0025] Embodied intelligent control center:

[0026] Issue task scheduling, state monitoring and business logic to embodied intelligence.

[0027] Pass the business intent to the network control service.

[0028] Further,

[0029] Step 1: Intent declaration

[0030] The control center plans to have embodied intelligence A and B perform collaborative transportation. It sends instructions to the embodied intelligence controller through the API.

[0031] Step 2: Prediction and decision

[0032] The state collector of the network control service continuously receives the health status of all links.

[0033] The AI prediction engine analyzes and concludes: "The currently dominant WiFi link may rise from 15ms to 80ms in the next 2 minutes, exceeding the task requirements; while the 5G link currently has a delay of 20ms and will remain stable at around 25ms in the next 10 minutes."

[0034] Based on this prediction, the strategy engine immediately makes a decision: "Switch the path of collaborative communication flow between embodied intelligence A and B from WiFi to 5G link".

[0035] Step 3: Strategy issuance and execution

[0036] The configuration distributor immediately issues the new path strategy to the clients of embodied intelligence A and B. After receiving the strategy, the client immediately adjusts the local routing table to send all traffic marked as high priority to the other IP through the 5G module.

[0037] Step 4: Continuous optimization and guarantee

[0038] During the entire task, the state collector continuously monitors the actual performance of the 5G link. If the 5G link suddenly experiences jitter (such as being blocked), the prediction engine quickly identifies this change, and the strategy engine can make a new decision based on the current state of all links.

[0039] After the task is completed, the control center notifies the controller, which automatically removes the temporary strategy and restores the default routing.

[0040] The beneficial effects of the present invention are

[0041] The present invention effectively solves the core pain points of existing technologies that cannot balance network determinism and business intelligence by deeply integrating artificial intelligence technology and software-defined networks. The present invention not only brings a qualitative leap in technical performance, but also produces significant positive effects in economic benefits, operational efficiency, and security, providing crucial network infrastructure support for the development of remote and cloud-based intelligent manufacturing. BRIEF DESCRIPTION OF DRAWINGS

[0042] Figure 1 is a data flow diagram working schematic of the present invention. DETAILED DESCRIPTION

[0043] In order to make the purposes, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0044] The present application provides an AI-driven adaptive network construction method for sensing services, predicting changes and dynamically adjusting, which ensures the reliability of embodied intelligent communication. By using an artificial intelligence algorithm LSTM (Long Short Term Memory Network) to predict network status, and then converting service intention into executable network strategy according to the demand of embodied intelligent service, the network path is dynamically optimized to provide the optimal and deterministic network path for embodied intelligent communication.

[0045] System modules:

[0046] Network control service, mainly for managing multiple terminal devices, monitoring network status, executing network strategy, sensing services and the like.

[0047] Embodied intelligent terminal communication client, with WiFi+5G module and functions such as dual-transmission selection and link aggregation, can execute the network control strategy issued by the embodied intelligent control center.

[0048] Embodied intelligent control center, responsible for task scheduling, state monitoring and service logic of all embodied intelligent terminals. It transmits "service intention" to the network control service, which is the caller of network service.

[0049] Network control service:

[0050] Gathers network and service status data reported by all terminals to construct global real-time network view and service view. A machine learning time series prediction model (LSTM (Long Short Term Memory Network)) is used to input historical and real-time link performance data (delay, jitter, packet loss), time (whether it is a working day), network load and the like. The performance index probability distribution of each available WAN link (5G, WiFi) in a future time window (such as 5 minutes in the future) is predicted.

[0051] Receives service intention from the embodied intelligent control center, such as ("start collaborative carrying task, last for 20 minutes, require network delay <40ms")

[0052] Combines AI prediction results, real-time network status and service strategy to make multi-objective optimization decision and calculate the globally optimal traffic path scheme at the current time.

[0053] Real-time and reliable path strategy generated by decision-making is issued to embodied intelligent terminal communication device clients.

[0054] Embodied intelligent terminal communication clients:

[0055] Network status collection: Real-time monitoring and reporting of embodied intelligent local network indicators, including but not limited to: end-to-end delay, jitter, packet loss rate (sending probe packets to the controller), and the status of available network interfaces (5G signal strength, WiFi strength).

[0056] Receive and strictly execute path strategies from network controllers, send through specified network interfaces (such as 5G, WiFi, dual transmission and reception, etc.).

[0057] Embodied intelligent control center:

[0058] Issue task scheduling, state monitoring, and business logic to embodied intelligence.

[0059] Pass "business intent" to network control services.

[0060] Workflow:

[0061] Step one: intent declaration

[0062] The control center plans for embodied intelligences A and B to perform collaborative transportation. It sends instructions to the embodied intelligence controller through API.

[0063] Step two: prediction and decision-making

[0064] The network control service's state collector continuously receives the health status of all links.

[0065] The AI prediction engine analyzes and concludes: "The currently dominant WiFi link may rise from 15ms to 80ms in the next 2 minutes, exceeding the task requirements; while the 5G link currently has a delay of 20ms and will remain stable at around 25ms for the next 10 minutes."

[0066] The strategy engine makes a decision based on this prediction: "Switch the path of collaborative communication flow between embodied intelligences A and B from WiFi to 5G link."

[0067] Step three: policy issuance and execution

[0068] The configuration distributor immediately issues the new path strategy to the clients of embodied intelligences A and B. After receiving the strategy, the client adjusts the local routing table to send all traffic marked as high priority to the other IP through the 5G module.

[0069] Step four: continuous optimization and guarantee

[0070] During the whole task, the state collector continuously monitors the actual performance of the 5G link. If the 5G link suddenly appears jitter (such as the signal is blocked), the prediction engine quickly identifies the change, and the strategy engine can make a decision again according to the current state of all links.

[0071] After the task is completed, the control center notifies the controller, and the controller automatically releases the temporary strategy, and the traffic returns to the default route.

[0072] The above merely describes the preferred embodiments of the present application, which are only used to explain the technical solutions of the present application, and are not used to limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. An adaptive network construction method for embodied intelligence, characterized in that, By using an artificial intelligence model to predict network status, and then converting business intent into executable network policies according to the business needs of embodied intelligence, dynamically optimizing network paths, and providing the best and deterministic network paths for embodied intelligence.

2. The method of claim 1, characterized in that, By using the business semantic information and network status information of the robot control center, an artificial intelligence model is used to predict future network status, and based on the prediction results and business policies, the optimal network path is dynamically selected for different business flows.

3. The method of claim 1, characterized in that, The business semantic information includes robot task type, data flow priority, and maximum delay requirement.

4. The method of claim 1, characterized in that, The artificial intelligence model uses a time series prediction model LSTM to predict the delay, jitter, and packet loss rate of WAN links.

5. The method of claim 1, characterized in that, It includes the following modules: Network control service module, managing multiple terminal devices, monitoring network status, executing network policies, and sensing business; Embodied intelligent terminal communication client, with WiFi+5G module, and with the functions of dual-transmitting and single-receiving and link aggregation, can execute the network control policies issued by the embodied intelligent control center; Embodied intelligent control center, issuing task scheduling, state monitoring, and business logic for all embodied intelligence.

6. The method of claim 5, characterized in that, Network control service module: Gathers all network and business status data reported by terminals, constructs global real-time network view and business view, uses machine learning time series prediction model, inputs historical and real-time link performance data, time, and network load, and predicts the performance index probability distribution of each available WAN link in the future time window; Receives business intent from the embodied intelligent control center; Integrates AI prediction results, real-time network status, and business policies for multi-objective optimization decision-making, and calculates the globally optimal traffic path scheme at the current time; Real-time and reliable path strategy generated by decision-making is issued to the embodied intelligent terminal communication device client.

7. The method of claim 5, characterized in that, Embodied intelligent terminal communication client: Network status acquisition: real-time monitoring and reporting of local network indicators of embodied intelligence, including end-to-end delay, jitter, packet loss rate, and status of available network interfaces; Receives and strictly executes path strategies from the network controller and sends through the specified network interface.

8. The method of claim 5, characterized in that, Embodied intelligent control center: Issues task scheduling, state monitoring, and business logic to embodied intelligence. Transfers business intent to the network control service.

9. The method of claim 5, characterized in that, Workflow: Step 1: Intent declaration The control center plans to have embodied intelligence perform collaborative transportation; it sends instructions to the embodied intelligent controller through API; Step 2: Prediction and decision-making The state collector of the network control service continuously receives the health status of all links; The AI prediction engine analyzes the data and draws a conclusion; The policy engine makes a decision based on the prediction. Step 3: Policy distribution and execution The configuration distributor immediately distributes the new path policy to the intelligent client; after receiving the policy, the client adjusts the local routing table and sends all traffic marked as high priority to the opposite IP through the 5G module. Step 4: Continuous optimization and guarantee During the entire task, the state collector continuously monitors the actual performance of the 5G link; if the 5G link suddenly experiences jitter, the prediction engine quickly identifies the change, and the policy engine can make a new decision based on the current state of all links. After the task is completed, the control center notifies the controller, which automatically removes the temporary policy and restores the default routing.

Images