Search

CN-121979685-A - Intelligent gateway equipment based on AI edge calculation

CN121979685ACN 121979685 ACN121979685 ACN 121979685ACN-121979685-A

Abstract

The invention belongs to the technical field of Internet of things, and discloses intelligent gateway equipment based on AI edge calculation, the system comprises a main processor, an AI acceleration processor, an adaptive task scheduling engine, a heterogeneous resource dynamic arrangement module, an adaptive power consumption management module and a two-way model verification and execution module which runs in a hardware trusted execution environment. The system comprises a task scheduling engine, a resource scheduling module, a power consumption management module, a security module and a security module, wherein the task scheduling engine queues according to task priority, the resource scheduling module accurately schedules tasks to a CPU or an AI processor according to task characteristics, the power consumption management module dynamically adjusts power consumption according to the load state of a task queue, and the security module performs double verification of sources and functions on an AI model before executing the tasks. The invention realizes the high real-time processing performance, the extremely high energy efficiency ratio and the solid safety guarantee of the intelligent gateway equipment through the deep coordination and the closed-loop control of the four modules.

Inventors

  • HU TAO
  • LIU XIAOLIN
  • Zhang Laien

Assignees

  • 苏州稳联科技有限公司

Dates

Publication Date
20260505
Application Date
20260403

Claims (10)

  1. 1. An intelligent gateway device based on AI edge computation, comprising: A main processor and an AI acceleration processor; an adaptive task scheduling engine configured to identify the type and priority of an incoming task according to preset rules and to assign the task to one of a plurality of priority queues; The heterogeneous resource dynamic scheduling module is connected with the adaptive task scheduling engine and is configured to acquire tasks from the plurality of priority queues and schedule the tasks to the main processor or the AI acceleration processor for execution according to the calculation characteristics of the tasks; An adaptive power consumption management module, connected to the adaptive task scheduling engine, configured to dynamically adjust a power consumption mode of at least one of the main processor and the AI-accelerated processor according to a real-time load state of the plurality of priority queues; And the two-way model verification and execution module is operated in a hardware trusted execution environment and is configured to perform two-way verification on the AI model before the AI acceleration processor executes the AI model.
  2. 2. The intelligent gateway device of claim 1, wherein the plurality of priority queues includes a real-time queue for handling real-time AI reasoning tasks, a near-real-time queue for handling data preprocessing or result aggregation, and a non-real-time queue for handling data reporting or system logging.
  3. 3. The intelligent gateway device of claim 1 or 2, wherein the heterogeneous resource dynamic orchestration module is specifically configured to schedule AI model reasoning tasks determined to require massively parallel computation to the AI acceleration processor for execution, and to schedule tasks determined to involve complex logic control or serial processing to the host processor for execution.
  4. 4. The intelligent gateway device of claim 2, wherein the adaptive power consumption management module is specifically configured to switch the power consumption mode of the AI acceleration processor to a high performance mode when the load of the real-time queue is above a first preset threshold and switch the main processor and the AI acceleration processor to a low power consumption mode when the load of all priority queues is below a second preset threshold.
  5. 5. The intelligent gateway device of claim 1, wherein the hardware trusted execution environment is a secure area built based on processor hardware isolation technology.
  6. 6. The intelligent gateway device of claim 1 or 5, wherein the two-way authentication comprises a first-way authentication to authenticate a digital signature of the AI model and a second-way authentication to perform a functional reasoning test on the AI model using preset security test data within the hardware trusted execution environment, the AI model being allowed to execute only after both the first-way authentication and the second-way authentication pass.
  7. 7. The intelligent gateway device of claim 2, wherein the adaptive task scheduling engine and the heterogeneous resource dynamic orchestration module work cooperatively to ensure that tasks in the real-time queue are scheduled preferentially over tasks in other queues.
  8. 8. The intelligent gateway device of claim 1, wherein the adaptive task scheduling engine is configured to periodically or event-triggered notify the adaptive power consumption management module of load status information of the plurality of priority queues.
  9. 9. The intelligent gateway device of claim 1, further comprising at least one network interface, at least one RS-485 interface, and at least one switching value input interface for collecting data from an external device or network as a source of the incoming tasks.
  10. 10. The intelligent gateway device of claim 1, wherein the output information of the adaptive task scheduling engine is used as input to the heterogeneous resource dynamic orchestration module and the adaptive power consumption management module, and execution of all AI models is managed by the two-way model verification and execution module, thereby forming an integrated closed-loop processing system.

Description

Intelligent gateway equipment based on AI edge calculation Technical Field The invention belongs to the technical field of the Internet of things, and particularly relates to intelligent gateway equipment based on AI edge calculation. Background With the rapid development of internet of things (IoT) and Artificial Intelligence (AI) technologies, the edge computing gateway serves as a core hub for connecting the physical world terminal device and the cloud digital world, and the functional and performance requirements of the edge computing gateway are increasingly higher. Modern edge computing gateways are no longer merely data transfer sites, but are given powerful local computing capabilities, particularly by integrating high-performance main processors (CPUs) and dedicated AI acceleration processors (e.g., TPU, NPUs), enabling them to perform complex AI reasoning tasks near the data source. The architecture can effectively reduce network delay, save transmission bandwidth and enhance data privacy. In the process of implementing the invention, the inventor finds that the existing edge computing gateway still exposes a plurality of technical bottlenecks in practical application: 1. the CPU and AI acceleration processors inside the gateway are typical heterogeneous computing resources, but most systems lack a sophisticated co-scheduling mechanism. When multiple tasks such as AI reasoning, data conversion, network communication and the like are concurrent, response delay of time-sensitive key AI tasks (such as real-time quality inspection on a production line) is often caused by indiscriminate resource contention, and the harsh requirements of industrial scenes cannot be met. 2. The traditional power consumption management strategy is mainly based on temperature or general CPU utilization rate for passive adjustment, and cannot be accurately matched with the actual peak and trough of the service load. This results in the gateway still running at higher power consumption during idle traffic periods, resulting in energy waste, and during peak traffic periods, the full peak performance of the hardware may not be realized due to simple power wall limitations. 3. AI models on edge gateways are typically remotely issued and updated by the cloud. In this process, the model may be at risk of being tampered with or replaced. It is difficult to protect against more complex attacks (such as model poisoning attacks) by means of conventional digital signature verification alone. Once a malicious AI model is deployed, it may lead to erroneous production decisions, system paralysis, or even security incidents with serious consequences. Disclosure of Invention The invention aims to solve the problems of low resource scheduling efficiency, mismatching of power consumption and performance and insufficient safety and reliability of AI model execution in the prior art at least to a certain extent. Therefore, the invention aims to provide an intelligent gateway device based on AI edge calculation, which can intelligently cooperate with internal heterogeneous resources, realize dynamic optimal balance of performance and power consumption, and establish a full-link trusted environment from model issuing to execution. The invention provides intelligent gateway equipment based on AI edge calculation, which comprises a main processor, an AI acceleration processor, an adaptive task scheduling engine, a heterogeneous resource dynamic scheduling module, an adaptive power consumption management module and a two-way model verification and execution module, wherein the adaptive task scheduling engine is configured to identify the type and the priority of an incoming task according to preset rules and distribute the task to one of a plurality of priority queues, the heterogeneous resource dynamic scheduling module is connected with the adaptive task scheduling engine and configured to acquire the task from the plurality of priority queues and schedule the task to the main processor or the AI acceleration processor for execution according to the calculation characteristics of the task, the adaptive power consumption management module is connected with the adaptive task scheduling engine and configured to dynamically adjust the power consumption mode of at least one of the main processor and the AI acceleration processor according to the real-time load state of the plurality of priority queues, and the two-way model verification and execution module is operated in a hardware trusted execution environment and configured to verify the AI model before the AI acceleration processor executes the AI model. Further, the plurality of priority queues include a real-time queue for processing real-time AI reasoning tasks, a near-real-time queue for processing data preprocessing or result aggregation, and a non-real-time queue for processing data reporting or system logging. Further, the heterogeneous resource dynamic arrangement module is specifically co