An edge AI computing device based on a 5G RedCap module

Abstract

This utility model relates to the field of edge AI device technology, specifically to an edge AI computing device based on a 5G RedCap module. It includes a main control unit for running lightweight AI models to achieve local inference and low-latency wide-area communication, a power management module, a communication unit for communicating with a corresponding cloud platform, a sensor management module for connecting corresponding sensors, and a file management module for caching AI models and data. This utility model achieves an integrated design for the edge AI computing device, possessing remote communication and local inference capabilities. It features rich interfaces for easy use and future upgrades, with tightly connected and mutually cooperating internal components, resulting in high integration. This reduces the number of peripheral modules, improves the stability and maintenance convenience of the edge AI computing device, and enhances deployment flexibility while maintaining low power consumption. It is applicable to various edge intelligence fields and has good versatility.

Description

Technical Field

[0001] This utility model relates to the field of edge AI device technology, specifically to an edge AI computing device based on a 5G RedCap module. Background Technology

[0002] With the development of artificial intelligence and the Internet of Things, terminal devices are placing higher demands on their "computing + communication" capabilities, especially in scenarios such as industrial sites, smart cities, and intelligent monitoring. These devices need to have local AI inference capabilities while also being able to stably connect to wide area networks for remote data interaction. Traditional solutions typically distribute AI inference and communication functions across multiple modules. The communication part is often implemented through 4G or Wi-Fi, while the AI ​​inference part relies on a high-power processing platform. This results in high hardware redundancy, requiring additional main control chips and AI acceleration chips. The components are separated and independent of each other. This architecture leads to high device costs, complex development, and is not conducive to rapid development and deployment. Furthermore, it has high power consumption, making it difficult to meet the needs of low-power edge devices. Therefore, it is necessary to develop a highly integrated edge AI computing device. Utility Model Content

[0003] The technical problem to be solved by this utility model is that existing edge devices generally distribute AI inference and communication functions in multiple modules, with each part being separate and independent of the others. This architecture results in high device costs, complex development, and is not conducive to rapid development and deployment. In addition, it has high power consumption, making it difficult to meet the requirements of low-power edge devices.

[0004] To solve the above-mentioned technical problems, this utility model adopts the following technical solution: an edge AI computing device based on a 5G RedCap module, comprising a main control unit for running lightweight AI models to achieve local inference and low-latency wide-area communication, a power management module, a communication unit for communicating with the corresponding cloud platform, a sensor management module for connecting the corresponding sensors, and a file management module for caching AI models and data. The main control unit is connected to the communication unit and communicates with the corresponding cloud platform through the communication unit. The main control unit is connected to the sensor management module and communicates with the corresponding sensors through the sensor management module. The main control unit is connected to the file management module. The power management module supplies power to the main control unit, the communication unit, the sensor management module, and the file management module.

[0005] When in operation, this utility model can realize the integrated design of edge AI computing device, with remote communication and local inference functions, and has rich interfaces for easy use and later upgrades. The internal parts are closely connected and cooperate with each other, with a high degree of integration, reducing peripheral modules, improving the stability and maintenance convenience of edge AI computing device, and improving deployment flexibility while ensuring low power consumption. It is applicable to a variety of edge intelligence fields and has good versatility.

[0006] Preferably, the main control unit is deployed with the TensorFlow Lite Micro framework, and the main control unit outputs inference results after receiving and processing the data transmitted by the sensor management module.

[0007] Preferably, the communication unit includes an RF integrated circuit and an RF front-end circuit. The main control unit is provided with an RF MIPI interface, an RF CTRL interface, and an IQ interface. The RF integrated circuit is connected to the main control unit through the RF MIPI interface, RF CTRL interface, and IQ interface, respectively. The RF integrated circuit is provided with a TX interface, a CTRL interface, a PRX interface, and a DRX interface. The RF front-end circuit is connected to the RF integrated circuit through the TX interface, CTRL interface, PRX interface, and DRX interface. The RF front-end circuit is provided with an ANT_MAIN interface and an ANT_DIV interface. The RF front-end circuit communicates with the corresponding cloud platform through the ANT_MAIN interface and the ANT_DIV interface.

[0008] Preferably, the communication unit further includes an auxiliary radio frequency power management module. The main control unit is connected to the radio frequency front-end circuit through the auxiliary radio frequency power management module to optimize and adjust the power supply and power distribution of the radio frequency front-end circuit. The auxiliary radio frequency power management module is connected to the corresponding cloud platform to establish a feedback adjustment link.

[0009] Preferably, the main control unit is equipped with an SPMI interface, and the main control unit is connected to the power management module through the SPMI interface to coordinate power supply and manage the power of each part. The power management module is equipped with a VDD_EXT interface and a VBAT_BB interface, and the power management module is connected to the corresponding cloud platform through the VDD_EXT interface and the VBAT_BB interface.

[0010] Preferably, the power management module has several interfaces, including at least one of the following: STATUS interface, PWRKEY interface, RESET_N interface, ADCs interface, and WLAN_SLP_CLK interface.

[0011] Preferably, the file management module is configured as an SLC NAND, the main control unit is equipped with an EBI / QSPI interface, and the file management module is connected to the main control unit through the EBI / QSPI interface.

[0012] Preferably, the main control unit is provided with an internal power management sub-module for realizing dynamic power consumption adjustment, and the power management module is connected to the main control unit through the internal power management sub-module.

[0013] Preferably, the main control unit is provided with several interfaces, including at least one of the following: B_Code interface, 1PPS interface, RGMII interface, USB2.0 interface, USIM interface, PCM interface, I2C interface, UART interface, GPIOs interface, SPI interface, PCIE interface, SDIO interface, and WLAN CTRL interface.

[0014] The beneficial technical effects of this utility model include:

[0015] This invention enables an integrated design of an edge AI computing device, featuring remote communication and local inference capabilities. It also boasts a wealth of interfaces for ease of use and future upgrades. The internal components are tightly connected and work together seamlessly, resulting in a high degree of integration. This reduces the number of peripheral modules, enhancing the stability and ease of maintenance of the edge AI computing device. It can improve deployment flexibility while ensuring low power consumption, making it suitable for various edge intelligence fields and demonstrating good versatility.

[0016] Other features and advantages of this utility model will be disclosed in detail in the following specific embodiments and accompanying drawings. Attached Figure Description

[0017] The present invention will be further described below with reference to the accompanying drawings:

[0018] Figure 1 A schematic diagram of the structure of an edge AI computing device based on a 5G RedCap module;

[0019] Figure 2 The circuit structure of an edge AI computing device based on a 5G RedCap module is shown in section 2. Detailed Implementation

[0020] The technical solutions of the present utility model will be explained and described below with reference to the accompanying drawings. However, the following embodiments are only preferred embodiments of the present utility model and not all of them. Other embodiments obtained by those skilled in the art based on the embodiments in the implementation methods without creative effort are all within the protection scope of the present utility model.

[0021] In the following description, terms such as “inner,” “outer,” “upper,” “lower,” “left,” and “right” are used only to facilitate the description of the embodiments and simplify the description, and are not intended to indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this utility model.

[0022] Please see Figure 1 This embodiment discloses an edge AI computing device based on a 5G RedCap module, including a main control unit 1 for running lightweight AI models to achieve local inference and low-latency wide-area communication, a power management module 2, a communication unit 3 for communicating with the corresponding cloud platform, a sensor management module 4 for connecting the corresponding sensors, and a file management module 5 for caching AI models and data. The following is a detailed description with reference to the accompanying drawings.

[0023] Please see Figure 1 and Figure 2 In this embodiment, the main control unit 1 can be configured as an ARM Cortex-A series processor. The processor runs a lightweight AI model, realizing an integrated design of local inference and remote communication. The main control unit 1 is connected to the communication unit 3 and communicates with the corresponding cloud platform through the communication unit 3. The main control unit 1 is connected to the sensor management module 4 and communicates with the corresponding sensors, such as cameras, temperature and humidity sensors, infrared sensors, etc., through the sensor management module 4. The main control unit 1 is connected to the file management module 5. The power management module 2 supplies power to the main control unit 1, the communication unit 3, the sensor management module 4, and the file management module 5.

[0024] When this embodiment is in operation, it can realize the integrated design of edge AI computing device, with remote communication and local inference functions, and has rich interfaces for easy use and later upgrades. The internal parts are closely connected and cooperate with each other, with a high degree of integration, reducing peripheral modules, improving the stability and maintenance convenience of edge AI computing device, and improving deployment flexibility while ensuring low power consumption. It is suitable for a variety of edge intelligence fields and has good versatility.

[0025] Preferably, the main control unit 1 is equipped with the TensorFlow Lite Micro framework. After receiving and processing the data transmitted by the sensor management module 4, the main control unit 1 outputs the inference results. In specific implementation, it is also necessary to lightweight the AI ​​model, optimize the model structure through integer quantization and pruning techniques, simplify and adapt to the Redcap platform, and deploy the AI ​​model in the edge AI computing device using the TensorFlow Lite Micro framework to achieve fast inference locally, such as image classification and anomaly detection.

[0026] In specific implementation, the communication unit 3 includes an RF integrated circuit and an RF front-end circuit. The main control unit 1 is equipped with an RF MIPI interface, an RF CTRL interface, and an IQ interface. The RF integrated circuit is connected to the main control unit 1 through the RF MIPI interface, RF CTRL interface, and IQ interface, respectively. The RF integrated circuit is equipped with a TX interface, a CTRL interface, a PRX interface, and a DRX interface. The RF front-end circuit is connected to the RF integrated circuit through the TX interface, CTRL interface, PRX interface, and DRX interface. The RF front-end circuit is equipped with an ANT_MAIN interface and an ANT_DIV interface. The RF front-end circuit communicates with the corresponding cloud platform through the ANT_MAIN interface and the ANT_DIV interface. Preferably, a modular communication unit 3 can be adopted to facilitate adaptive adjustment according to the actual working scenario. For example, it can be adapted to various application scenarios through 5G or IoT communication modules.

[0027] As a further improvement of this embodiment, the communication unit 3 also includes an auxiliary RF power management module. The main control unit 1 is connected to the RF front-end circuit through the auxiliary RF power management module to optimize and adjust the power supply and power distribution of the RF front-end circuit. The auxiliary RF power management module is connected to the corresponding cloud platform to establish a feedback adjustment link. When the signal strength is lower than the threshold, the main control unit 1 increases the gain value of the PA by sending a command to enhance the transmission power, up to +3dBm. After the signal is restored, the power is automatically reduced to avoid energy waste. It can work with the RF power management module and the RF front-end circuit to optimize the power supply and signal amplification of the RF link, improve the wireless transmission performance, and achieve adaptive dynamic voltage adjustment through the feedback adjustment link, thereby reducing the overall power consumption.

[0028] In practical implementation, the main control unit 1 is equipped with an SPMI interface. The main control unit 1 is connected to the power management module 2 through the SPMI interface to coordinate power supply and manage the power of each part. The power management module 2 is equipped with a VDD_EXT interface and a VBAT_BB interface. The power management module 2 is connected to the corresponding cloud platform through the VDD_EXT interface and the VBAT_BB interface. At the same time, in order to further optimize energy consumption, the main control unit 1 is equipped with an internal power management sub-module for realizing dynamic power consumption adjustment. The power management module 2 is connected to the main control unit 1 through the internal power management sub-module.

[0029] Preferably, the power management module 2 is provided with several interfaces, including at least one of the following: STATUS interface, PWRKEY interface, RESET_N interface, ADCs interface, and WLAN_SLP_CLK interface. When in operation, it is suitable for various working needs and has high compatibility.

[0030] In practical implementation, the file management module 5 is set to SLC NAND, such as 1Gbit flash memory, and the main control unit 1 is equipped with an EBI / QSPI interface. The file management module 5 is connected to the main control unit 1 through the EBI / QSPI interface.

[0031] Preferably, to be suitable for various working scenarios, the main control unit 1 is equipped with several interfaces, including at least one of the following: B_Code interface, 1PPS interface, RGMII interface, USB2.0 interface, USIM interface, PCM interface, I2C interface, UART interface, GPIOs interface, SPI interface, PCIE interface, SDIO interface, and WLAN CTRL interface. These interfaces can be customized and called as needed in actual work.

[0032] The beneficial technical effects of this embodiment include: This utility model can realize the integrated design of edge AI computing device, with remote communication and local inference functions, and has rich interfaces for easy use and later upgrades. The internal parts are closely connected and cooperate with each other, with a high degree of integration, reducing peripheral modules, improving the stability and maintenance convenience of edge AI computing device, and improving deployment flexibility while ensuring low power consumption. It is applicable to a variety of edge intelligence fields and has good versatility.

[0033] The above description is merely a specific embodiment of this utility model, but the protection scope of this utility model is not limited thereto. Those skilled in the art should understand that this utility model includes, but is not limited to, the content described in the accompanying drawings and the specific embodiments above. Any modifications that do not depart from the functional and structural principles of this utility model will be included within the scope of the claims.

Claims

1. An edge AI computing device based on 5G RedCap module, characterized in that: The system includes a main control unit (1) for running lightweight AI models to achieve local inference and low-latency wide-area communication, a power management module (2), a communication unit (3) for communicating with the corresponding cloud platform, a sensor management module (4) for connecting the corresponding sensors, and a file management module (5) for caching AI models and data. The main control unit (1) is connected to the communication unit (3) and communicates with the corresponding cloud platform through the communication unit (3). The main control unit (1) is connected to the sensor management module (4) and communicates with the corresponding sensors through the sensor management module (4). The main control unit (1) is connected to the file management module (5). The power management module (2) supplies power to the main control unit (1), the communication unit (3), the sensor management module (4), and the file management module (5).

2. The edge AI computing device based on the 5G RedCap module according to claim 1, characterized in that: The main control unit (1) is equipped with the TensorFlow Lite Micro framework. After receiving and processing the data transmitted by the sensor management module (4), the main control unit (1) outputs the inference result.

3. The edge AI computing device based on the 5G RedCap module according to claim 1, characterized in that: The communication unit (3) includes a radio frequency integrated circuit and a radio frequency front-end circuit. The main control unit (1) is provided with an RF MIPI interface, an RF CTRL interface and an IQ interface. The radio frequency integrated circuit is connected to the main control unit (1) through the RF MIPI interface, the RF CTRL interface and the IQ interface respectively. The radio frequency integrated circuit is provided with a TX interface, a CTRL interface, a PRX interface and a DRX interface. The radio frequency front-end circuit is connected to the radio frequency integrated circuit through the TX interface, the CTRL interface, the PRX interface and the DRX interface. The radio frequency front-end circuit is provided with an ANT_MAIN interface and an ANT_DIV interface. The radio frequency front-end circuit communicates with the corresponding cloud platform through the ANT_MAIN interface and the ANT_DIV interface.

4. The edge AI computing device based on a 5G RedCap module according to claim 1, characterized in that: The communication unit (3) also includes an auxiliary radio frequency power management module. The main control unit (1) is connected to the radio frequency front-end circuit through the auxiliary radio frequency power management module to optimize and adjust the power supply and power distribution of the radio frequency front-end circuit. The auxiliary radio frequency power management module is connected to the corresponding cloud platform to establish a feedback adjustment link.

5. An edge AI computing device based on a 5G RedCap module according to claim 1, characterized in that: The main control unit (1) is equipped with an SPMI interface. The main control unit (1) is connected to the power management module (2) through the SPMI interface to coordinate power supply and manage the power of each part. The power management module (2) is equipped with a VDD_EXT interface and a VBAT_BB interface. The power management module (2) is connected to the corresponding cloud platform through the VDD_EXT interface and the VBAT_BB interface.

6. The edge AI computing device based on a 5G RedCap module according to claim 1, characterized in that: The power management module (2) has several interfaces, including at least one of the following: STATUS interface, PWRKEY interface, RESET_N interface, ADCs interface and WLAN_SLP_CLK interface.

7. An edge AI computing device based on a 5G RedCap module according to claim 1, characterized in that: The file management module (5) is configured as an SLC NAND, and the main control unit (1) is equipped with an EBI / QSPI interface. The file management module (5) is connected to the main control unit (1) through the EBI / QSPI interface.

8. An edge AI computing device based on a 5G RedCap module according to claim 1, characterized in that: The main control unit (1) is provided with an internal power management sub-module for realizing dynamic power consumption adjustment, and the power management module (2) is connected to the main control unit (1) through the internal power management sub-module.

9. An edge AI computing device based on a 5G RedCap module according to claim 1, characterized in that: The main control unit (1) is provided with several interfaces, including at least one of the following: B_Code interface, 1PPS interface, RGMII interface, USB2.0 interface, USIM interface, PCM interface, I2C interface, UART interface, GPIOs interface, SPI interface, PCIE interface, SDIO interface, and WLAN CTRL interface.

Images