An artificial intelligence (AI) communication method and apparatus

Abstract

The application provides an artificial intelligence (AI) communication method applied to any two devices in a communication system for communication, and the method comprises the following steps: a first device acquires AI capability information, wherein the AI capability information comprises a time and / or energy consumption of the first device for executing each reference AI unit in at least one reference AI unit; and the first device sends the AI capability information to a second device. Further, the method further comprises the following steps: the first device receives configuration information sent by the second device, wherein the configuration information indicates that the first device starts an AI communication mode, or the configuration information indicates at least one AI model, or the configuration information indicates configuration parameters of the at least one AI model, or the configuration information indicates an acquisition method of the at least one AI model. In this way, the second device can evaluate the AI capability of the first device and the matching condition of the AI model through the AI capability information reported by the first device, so that the simplicity, efficiency and accuracy of the evaluation result of the second device can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of communication, and particularly relates to an artificial intelligence (AI) communication method and device. BACKGROUND

[0002] Artificial intelligence (AI) technology is a branch of computer science, which runs through the history of computer development, and is an important development direction of the information technology industry. With the development of communication technology, more and more applications will be intelligentized through AI. At present, AI technology is introduced into a wireless communication system, and AI modules may be gradually used to replace functional modules in the wireless communication system. After the wireless communication system introduces AI technology, a possible working mode is that a network device sends an AI model to a terminal device, and the terminal device receives the AI model from the network device and applies the AI model for wireless communication.

[0003] Different terminal devices have different AI capabilities, so the AI model sent by the network device may not be executed by the terminal device, which may result in that the AI model cannot be applied for wireless communication. SUMMARY

[0004] The present application provides an AI communication method and device, so as to better apply AI technology in a wireless communication system.

[0005] In a first aspect, the present application provides an AI communication method, which can be executed by a first device. The first device can be a terminal device, or a receiver of an AI model in a communication system. The method can be implemented by the following steps: the first device obtains AI capability information, the AI capability information including time and / or energy consumption of the first device for executing each reference AI unit in at least one reference AI unit; and the first device sends the AI capability information to a second device.

[0006] In the implementation mode, the second device can evaluate the matching of the AI model and the AI capability of the first device by sending the AI capability information of the first device to the second device, so as to ensure the feasibility of the first device and the second device using the AI model for communication.

[0007] In a possible design, after the first device sends the AI capability information to the second device, the method further includes: the first device receives configuration information sent by the second device, wherein the configuration information indicates that the first device starts an AI communication mode, or the configuration information indicates at least one AI model, or the configuration information indicates configuration parameters of at least one AI model, or the configuration information indicates an acquisition method of at least one AI model.

[0008] In a possible design, the at least one AI model is determined according to the AI capability information.

[0009] In a possible design, the configuration information is response information of the AI capability information.

[0010] In a possible design, the time for the first device to execute the first AI reference unit of the at least one reference AI unit is a first time value, a first time level, or a first time range, and the energy consumption of the first device for executing the first AI reference unit of the at least one reference AI unit is a first energy consumption value, a first energy consumption level, or a first energy consumption range.

[0011] In a possible design, the AI capability information includes a quantity of input data used by each reference AI unit of the at least one reference AI unit executed by the first device.

[0012] In a possible design, the AI capability information includes one or more of the following: a numerical precision of input data used by each reference AI unit of the at least one reference AI unit executed by the first device; a numerical precision of a weight of each reference AI unit of the at least one reference AI unit; and an operation precision of each reference AI unit of the at least one reference AI unit executed by the first device.

[0013] In a possible design, the resource used by each reference AI unit of the at least one reference AI unit executed by the first device is all available computing resources of the first device. Based on the AI capability information, the second device can more conveniently, efficiently, and accurately evaluate the matching of the AI capability of the first device and the AI model complexity, and thus better perform AI communication.

[0014] In a possible design, the AI capability information includes at least one of a time upper limit value, an energy consumption upper limit value, and a resource usage of the first device for executing the AI model. In this way, the first device can send at least one of the time upper limit value information, the energy consumption upper limit value information, and the resource usage information of the first device for executing the AI model to the second device, to inform the second device of the time budget requirement, the energy consumption limit, or the resource usage of the first device for executing the AI model.

[0015] In a possible design, before the first device sends the AI capability information to the second device, the method further includes: receiving, by the first device, request information from the second device, the request information being used to request the first device to send the AI capability information to the second device.

[0016] In a possible design, the first device sends the AI capability information to the second device in the following manners: the first device periodically sends the AI capability information to the second device; or the first device sends the AI capability information to the second device when the first device accesses a network in which the second device is located; or the first device sends the AI capability information to the second device when the first device establishes a communication connection with the second device; or the first device sends the AI capability information to the second device when a computing resource used by the first device to execute an AI model changes.

[0017] In a second aspect, the present application provides an AI communication method, which can be executed by a second device. The second device can be a network device or a sender of an AI model in a communication system. The method can be implemented through the following steps: the second device receives AI capability information sent by a first device, and the AI capability information includes time and / or energy consumption of the first device in executing at least one reference AI unit.

[0018] In a possible design, after the second device receives the AI capability information sent by the first device, the method further includes: the second device sends configuration information to the first device, where the configuration information indicates that the first device starts an AI communication mode, or the configuration information indicates at least one AI model, or the configuration information indicates configuration parameters of the at least one AI model, or the configuration information indicates an acquisition method of the at least one AI model, and the at least one AI model is determined according to the AI capability information.

[0019] In a possible design, the at least one AI model is determined according to the AI capability information.

[0020] In a possible design, the configuration information is response information of the AI capability information.

[0021] In a possible design, the time of the first device in executing a first AI reference unit in the at least one reference AI unit is a first time value, a first time level or a first time range, and the energy consumption of the first device in executing the first AI reference unit in the at least one reference AI unit is a first energy consumption value, a first energy consumption level or a first energy consumption range.

[0022] In a possible design, the AI capability information includes a quantity of input data used by each reference AI unit in the at least one reference AI unit.

[0023] In a possible design, the AI capability information includes one or more of the following: a numerical precision of input data used by each reference AI unit in the at least one reference AI unit; a numerical precision of a weight of each reference AI unit in the at least one reference AI unit; and an operation precision of each reference AI unit in the at least one reference AI unit.

[0024] In a possible design, the AI capability information includes at least one of a time upper limit value, an energy consumption upper limit value, and a resource usage of the first device for executing the AI model.

[0025] In a possible design, before the second device receives the AI capability information sent by the first device, the method further includes: sending, by the second device, request information to the first device, where the request information is used to request the first device to send the AI capability information to the second device.

[0026] In a possible design, when the configuration information indicates at least one AI model, or the configuration indicates configuration parameters of the at least one AI model, or the configuration information indicates an obtaining method of the at least one AI model, before the second device sends the configuration information to the first device, the method further includes: determining, by the second device, the at least one AI model according to the AI capability information.

[0027] In a possible design, determining, by the second device, the at least one AI model according to the AI capability information includes: determining, by the second device, the at least one AI model according to the AI capability information and M sets of similarity information corresponding to the M AI models, where each set of similarity information in the M sets of similarity information is similarity information of one AI model in the M AI models and K reference AI units. M and K are positive integers.

[0028] In a possible design, the M AI models are pre-stored in the second device or other devices; or the M AI models are obtained by the second device from other devices; or the M AI models are generated by the second device.

[0029] In a possible design, a first set of similarity information in the M sets of similarity information is similarity information of a first AI model and the K reference AI units, the first set of similarity information includes first similarity information, the first similarity information is associated with a first proportion and / or a second proportion, the first proportion is a proportion of a total number of layers same as a first reference AI unit in the first AI model in a total number of layers of the first AI model, the second proportion is a proportion of a total calculation amount of layers same as the first reference AI unit in the first AI model in a total calculation amount of the first AI model, the M AI models include the first AI model, and the K reference AI units include the first reference AI unit.

[0030] In another possible design, the first group of similarity information in the M groups of similarity information is similarity information of the first AI model and the K reference AI units, the first group of similarity information includes first similarity information, and the first similarity information is associated with a first proportion and / or a second proportion, where the first proportion is a proportion of a total number of layers in the first AI model that are same as a layer in the first reference AI unit to a total number of layers in the first AI model, and the second proportion is a proportion of a total amount of calculation in the first AI model that is same as an amount of calculation in the first reference AI unit to a total amount of calculation in the first AI model, the M AI models include the first AI model, and the K reference AI units include the first reference AI unit.

[0031] The beneficial effects of the second aspect and various possible designs can refer to the descriptions related to the first aspect, which will not be repeated here.

[0032] In a third aspect, the present application provides an AI communication method, which can be executed by a first device. The first device can be a terminal device or a receiver of an AI model in a communication system. The method can be implemented by the following steps: the first device receives AI model information sent by a second device, the AI model information including M groups of similarity information corresponding to M AI models, each group of similarity information in the M groups of similarity information being similarity information of one of the M AI models and K reference AI units, where M and K are positive integers; and the first device sends feedback information to the second device according to the AI model information.

[0033] In the implementation, the second device sends the AI model information to the first device, and the first device can evaluate the matching of the AI model and the AI capability of the first device, thereby ensuring the feasibility of the first device and the second device using the AI model to communicate.

[0034] In one possible design, the first group of similarity information in the M groups of similarity information is similarity information of the first AI model and the K reference AI units, the first group of similarity information includes first similarity information, and the first similarity information is associated with a first proportion and / or a second proportion, where the first proportion is a proportion of a total number of layers in the first AI model that are same as a layer in the first reference AI unit to a total number of layers in the first AI model, and the second proportion is a proportion of a total amount of calculation in the first AI model that is same as an amount of calculation in the first reference AI unit to a total amount of calculation in the first AI model, the M AI models include the first AI model, and the K reference AI units include the first reference AI unit.

[0035] In another possible design, the first similarity information in the M groups of similarity information is similarity information of the first AI model and K reference AI units, the first group of similarity information includes first similarity information, the first similarity information is associated with a first proportion and / or a second proportion, the first proportion is a proportion of a total number of units same as the first reference AI unit in the first AI model in a total number of units of the first AI model, and the second proportion is a proportion of a total amount of calculation of the units same as the first reference AI unit in the first AI model in a total amount of calculation of the first AI model, the M AI models include the first AI model, and the K reference AI units include the first reference AI unit.

[0036] In a possible design, the AI model information includes one or more of the following: a numerical precision of input data used when executing each of the M AI models; a numerical precision of a weight of each of the M AI models; and an operation precision when executing each of the M AI models.

[0037] In a possible design, the AI model information includes a total number of layers and / or a total amount of calculation of each of the M AI models.

[0038] Based on the AI model information, the first device can more conveniently, efficiently and accurately evaluate the matching of the AI capability of the first device and the AI model complexity, and thus better perform AI communication.

[0039] In a possible design, the AI model information includes a time upper limit value for executing the AI model. In this way, the first device can obtain a time budget requirement of the second device for the first device to execute the to-be-downloaded AI model, so as to better determine whether to start the AI communication mode or select a suitable AI model.

[0040] In a possible design, the feedback information indicates that the first device requests to start the AI mode, or the feedback information indicates an evaluation result of at least one AI model in the M AI models, or the feedback information requests the second device to send the at least one AI model to the first device, where the M AI models include the at least one AI model.

[0041] In a possible design, after the first device sends the feedback information to the second device according to the AI model information, the method further includes: receiving, by the first device, configuration information sent by the second device, the configuration information indicating that the first device starts the AI mode, or the configuration information indicating the at least one AI model, or the configuration information indicating a configuration parameter of the at least one AI model, or the configuration information indicating an obtaining method of the at least one model, where the M AI models include the at least one AI model.

[0042] In a possible design, before the first device receives the AI model information sent by the second device, the method further includes: sending, by the first device, request information to the second device, where the request information is used to request the second device to send the AI model information to the first device.

[0043] In a possible design, before the first device sends the feedback information to the second device based on the AI model information, the method further includes: determining, by the first device, the feedback information according to the AI model information and AI capability information of the first device.

[0044] In a possible design, the AI capability information of the first device indicates time and / or energy consumption of the first device for executing at least one reference AI unit. For possible designs of the AI capability information, refer to descriptions in the first aspect, which are not repeated here.

[0045] In a fourth aspect, the present application provides an AI communication method, which can be executed by a second device. The second device can be a network device, or a sender of an AI model in a communication system. The method can be implemented through the following steps: sending, by the second device, AI model information to a first device, where the AI model information includes M sets of similarity information corresponding to M AI models, and each set of similarity information is similarity information between one AI model and K reference AI units, where M and K are positive integers; and receiving, by the second device, feedback information sent by the first device, where the feedback information is determined according to the AI model information.

[0046] In a possible design, a first set of similarity information in the M sets of similarity information is similarity information between a first AI model and the K reference AI units, and the first set of similarity information includes first similarity information, where the first similarity information is associated with a first proportion and / or a second proportion, the first proportion is a proportion of a total number of layers identical to a first reference AI unit in the first AI model in a total number of layers of the first AI model, and the second proportion is a proportion of a total calculation amount of layers identical to the first reference AI unit in the first AI model in a total calculation amount of the first AI model, the M AI models include the first AI model, and the K reference AI units include the first reference AI unit.

[0047] In another possible design, the first similarity information in the M groups of similarity information is similarity information of the first AI model and K reference AI units, the first group of similarity information includes first similarity information, the first similarity information is associated with a first proportion and / or a second proportion, the first proportion is a proportion of a total number of units same as the first reference AI unit in the first AI model in a total number of units of the first AI model, and the second proportion is a proportion of a total amount of calculation of units same as the first reference AI unit in the first AI model in a total amount of calculation of the first AI model, the M AI models include the first AI model, and the K reference AI units include the first reference AI unit.

[0048] In a possible design, the AI model information includes one or more of the following: a numerical precision of input data used when executing each of the M AI models; a numerical precision of a weight of each of the M AI models; and an operation precision when executing each of the M AI models.

[0049] In a possible design, the AI model information includes a total number of layers and / or a total amount of calculation of each of the M AI models.

[0050] In a possible design, the AI model information includes an upper limit value of a time for executing the AI model.

[0051] In a possible design, the feedback information indicates that the first device requests to start the AI mode, or the feedback information indicates an evaluation result of at least one AI model in the M AI models, or the feedback information requests the second device to send the at least one AI model to the first device, where the M AI models include the at least one AI model.

[0052] In a possible design, after the second device receives the feedback information sent by the first device, the method further includes: the second device sends configuration information to the first device, the configuration information indicates that the first device starts the AI mode, or the configuration information indicates the at least one AI model, or the configuration information indicates a configuration parameter of the at least one AI model, or the configuration information indicates an obtaining method of the at least one AI model, where the M AI models include the at least one AI model.

[0053] In a possible design, before the second device sends the AI model information to the first device, the method further includes: the second device receives request information sent by the first device, the request information is used to request the second device to send the AI model information to the first device.

[0054] In a possible design, the M AI models are pre-stored in the second device or other devices, or the M AI models are obtained by the second device from other devices, or the M AI models are generated by the second device.

[0055] The beneficial effects of the fourth aspect and various possible designs can be referred to the description related to the third aspect, and thus will not be repeated here.

[0056] In the fifth aspect, the present application further provides a communication apparatus, which can be a terminal device, or the communication apparatus can be a receiving device in a communication system, and the communication apparatus has the functions of the first device in any of the above-mentioned first aspect or third aspect. The functions can be implemented by hardware, or by hardware executing corresponding software. The hardware or software includes one or more modules corresponding to the above-mentioned functions.

[0057] In a possible design, the communication apparatus includes a transceiver and a processing unit, which can perform the corresponding functions of the first device in any of the above-mentioned first aspect or third aspect, and details are referred to the description in the method examples, which will not be repeated here.

[0058] In a possible design, the communication apparatus includes a transceiver and a processor, and optionally includes a memory, where the transceiver is configured to transceive data and to communicate with other devices in a communication system, and the processor is configured to support the communication apparatus to perform the corresponding functions of the first device in any of the above-mentioned first aspect or third aspect. The memory is coupled to the processor, and stores necessary program instructions and data of the communication apparatus.

[0059] In the sixth aspect, the present application further provides a communication apparatus, which can be a network device, or the communication apparatus can be a transmitting device in a communication system, and the communication apparatus has the functions of the second device in any of the above-mentioned second aspect or fourth aspect. The functions can be implemented by hardware, or by hardware executing corresponding software. The hardware or software includes one or more modules corresponding to the above-mentioned functions.

[0060] In a possible design, the communication apparatus includes a transceiver and a processing unit, which can perform the corresponding functions of the second device in any of the above-mentioned second aspect or fourth aspect, and details are referred to the description in the method examples, which will not be repeated here.

[0061] In a possible design, the communication apparatus includes a transceiver and a processor, and optionally includes a memory, where the transceiver is configured to transceive data and to communicate with other devices in a communication system, and the processor is configured to support the communication apparatus to perform the corresponding functions of the second device in any of the above-mentioned second aspect or fourth aspect. The memory is coupled to the processor, and stores necessary program instructions and data of the communication apparatus.

[0062] In a seventh aspect, an embodiment of the present application provides a communication system, which can include the first device and the second device mentioned above.

[0063] In an eighth aspect, a computer readable storage medium is provided in an embodiment of the present application, which stores program instructions. When the program instructions are run on a computer, the computer is caused to perform the method in any one of the first aspect to the fourth aspect and any possible design thereof. For example, the computer readable storage medium can be any available medium that can be accessed by a computer. For example but not limited to: the computer readable medium can include a non-transitory computer readable medium, a random access memory (RAM), a read-only memory (ROM), an electrically EPROM (EEPROM), a CD-ROM or other optical disk storage, a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and can be accessed by a computer.

[0064] In a ninth aspect, a computer program product including computer program code or instructions is provided in an embodiment of the present application, which, when run on a computer, causes the computer to implement the method in any one of the first aspect to the fourth aspect and any possible design thereof.

[0065] In a tenth aspect, a chip is provided in the present application, which includes a processor coupled with a memory, for reading and executing program instructions stored in the memory, so as to cause the chip to implement the method in any one of the first aspect to the fourth aspect and any possible design thereof.

[0066] The technical effects of each of the fifth aspect to the tenth aspect and each possible design thereof can refer to the technical effect description of the various possible schemes of the first aspect to the fourth aspect, which will not be repeated here. BRIEF DESCRIPTION OF DRAWINGS

[0067] Figure 1 A schematic diagram of a communication system architecture is provided in the present application;

[0068] Figure 2 A schematic diagram of a process of wireless communication by a network device and a terminal device using an AI model is provided in an embodiment of the present application;

[0069] Figure 3 A schematic diagram of an AI communication method flow is provided in an embodiment of the present application;

[0070] Figure 4 Flowchart for AI capability and AI model matching degree evaluation of a network device in an embodiment of the present application;

[0071] Figure 5 Flowchart two for an AI communication method in an embodiment of the present application;

[0072] Figure 6 Flowchart for AI capability and AI model matching degree evaluation of a terminal device in an embodiment of the present application;

[0073] Figure 7 Structure diagram one of a communication apparatus in an embodiment of the present application;

[0074] Figure 8 Structure diagram two of a communication apparatus in an embodiment of the present application. DETAILED DESCRIPTION

[0075] The technical solutions in the present application will be described in further detail below with reference to the accompanying drawings.

[0076] The technical solutions provided by the present application can be applied to various communication systems, such as a 5th generation (5G) or new radio (NR) system, a long term evolution (LTE) system, an LTE frequency division duplex (FDD) system, an LTE time division duplex (TDD) system, etc. The technical solutions provided by the present application can also be applied to future communication systems, such as a 6th generation mobile communication system. The technical solutions provided by the present application can also be applied to device to device (D2D) communication, vehicle-to-everything (V2X) communication, machine to machine (M2M) communication, machine type communication (MTC), and internet of things (IoT) communication systems or other communication systems.

[0077] An AI communication method and apparatus are provided in embodiments of the present application. The method and apparatus described in the present application are based on the same technical concept. Since the principles of the method and apparatus for solving problems are similar, the implementation of the apparatus and the method can be mutually referred to, and the repeated parts will not be described again.

[0078] In the description of the present application, the words "first", "second", etc. are used only for the purpose of distinguishing the described objects, and cannot be understood as indicating or implying relative importance or indicating or implying an order.

[0079] In the description of the present application, "at least one" means one or more, and more than two means two or more.

[0080] In the description of the present application, unless otherwise specified, " / " means or, for example, A / B can mean A or B; "and / or" herein only describes the association relationship of the associated objects, which means that there can be three relationships, for example, A and / or B can mean that A exists alone, A and B exist together, and B exists alone. In addition, in the description of the embodiments of the present application, "multiple" means two or more. In order to more clearly describe the technical solutions of the embodiments of the present application, the downlink scheduling method and device provided by the embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0081] Figure 1 A schematic diagram of a wireless communication system suitable for the embodiments of the present application is shown. As shown in Figure 1 The wireless communication system can include at least one network device, such as Figure 1 The network device 111, the network device 112, the wireless communication system can also include at least one terminal device, such as Figure 1 The terminal device 121, the terminal device 122, the terminal device 123, the terminal device 124, the terminal device 125, the terminal device 126, the terminal device 127. The network device and the terminal device can be configured with multiple antennas, and the network device and the terminal device can communicate using multiple antenna technology.

[0082] Among them, when the network device and the terminal device communicate, the network device can manage one or more cells, and there can be an integer number of terminal devices in a cell. It should be noted that the cell can be understood as an area within the wireless signal coverage of the network device.

[0083] The present application can be used in the communication scenario between the network device and the terminal device, such as the network device 111 and the terminal device 121, the terminal device 122, the terminal device 123 can communicate; for example, the network device 111 and the network device 112 can communicate with the terminal device 124. The present application can also be used in the communication scenario between the terminal device and the terminal device, such as the terminal device 122 can communicate with the terminal device 125. The present application can also be used in the communication scenario between the network device and the network device, such as the network device 111 can communicate with the network device 112.

[0084] It should be understood that Figure 1The simplified schematic diagram is only for the convenience of understanding, and the application is not limited thereto. Embodiments of the application can be applied to any communication scenario of communication between a sending end device and a receiving end device.

[0085] The terminal device in the embodiments of the application can also be referred to as a user equipment (UE), an access terminal, a user unit, a user station, a mobile station, a mobile station, a remote station, a remote terminal, a mobile device, a user terminal, a terminal, a wireless communication device, a user agent or a user apparatus.

[0086] The terminal device can be a device providing voice / data to a user, for example, a handheld device with wireless connection function, a vehicle-mounted device, etc. At present, some examples of the terminal are: a mobile phone, a tablet computer, a notebook computer, a palm computer, a mobile internet device (MID), a wearable device, a virtual reality (VR) device, an augmented reality (AR) device, a wireless terminal in industrial control, a wireless terminal in self-driving, a wireless terminal in remote medical surgery, a wireless terminal in smart grid, a wireless terminal in transportation safety, a wireless terminal in smart city, a wireless terminal in smart home, a cellular phone, a cordless phone, a session initiation protocol (SIP) phone, a wireless local loop (WLL) station, a personal digital assistant (PDA), a handheld device with wireless communication function, a computing device or other processing device connected to a wireless modem, a wearable device, a terminal device in a 5G network or a terminal device in a future evolved public land mobile network (PLMN), etc., and the embodiments of the application are not limited thereto.

[0087] By way of example and not limitation, in embodiments of the present application, the terminal device can also be a wearable device. The wearable device can also be referred to as a wearable smart device, which is a general term for devices that are designed and developed by applying wearable technology to daily wear, such as glasses, gloves, watches, clothing, and shoes. The wearable device is a portable device that is directly worn on the body or integrated into the clothes or accessories of the user. The wearable device is not only a hardware device, but also has powerful functions through software support and data interaction and cloud interaction. The general wearable smart device includes devices with full functions and large sizes, which can realize complete or partial functions without relying on a smart phone, such as smart watches or smart glasses, and devices that focus on a certain type of application function and need to be used in cooperation with other devices, such as smart phones, such as various smart wristbands and smart jewelry for monitoring vital signs.

[0088] In embodiments of the present application, the device for implementing the function of the terminal device can be a terminal device, or a device capable of supporting the terminal device to implement the function, such as a chip system or a chip, which can be installed in the terminal device. In embodiments of the present application, the chip system can be composed of a chip, or can include a chip and other discrete devices.

[0089] The network device in the embodiments of the present application can be a device for communicating with a terminal device, and the network device can also be referred to as an access network device or a radio access network device, for example, the network device can be a base station. The network device in the embodiments of the present application can refer to a radio access network (RAN) node (or device) for accessing a terminal device to a wireless network. The base station can broadly cover various names in the following or be replaced by the following names, such as: Node B (NodeB), evolved Node B (eNB), next generation Node B (gNB), relay station, access point, transmitting and receiving point (TRP), transmitting point (TP), primary station, secondary station, motor slide retainer (MSR) node, home base station, network controller, access node, wireless node, access point (AP), transmission node, transceiver node, baseband unit (BBU), remote radio unit (RRU), active antenna unit (AAU), remote radio head (RRH), central unit (CU), distributed unit (DU), positioning node, etc. The base station can be a macro base station, a micro base station, a relay node, a donor node or the like, or a combination thereof. The base station can also refer to a communication module, modem or chip for being arranged in the foregoing device or apparatus. The base station can also be a mobile switching center and a device assuming a base station function in D2D, V2X, M2M communication, a network side device in 6G network, a device assuming a base station function in future communication system, etc. The base station can support networks of the same or different access technologies. The embodiments of the present application do not limit the specific technology and specific device form adopted by the network device.

[0090] The base station can be fixed or mobile. For example, a helicopter or a drone can be configured to act as a mobile base station, and one or more cells can move according to the location of the mobile base station. In other examples, the helicopter or the drone can be configured to serve as a device communicating with another base station.

[0091] In some deployments, the network device mentioned in the embodiments of the present application can be a device including a CU, or a DU, or a device including a CU and a DU, or a device including a control plane CU node (central unit-control plane, CU-CP) and a user plane CU node (central unit-user plane, CU-UP), and a DU node.

[0092] The network device and the terminal device can be deployed on land, including indoor or outdoor, handheld or vehicle-mounted; can also be deployed on water; can also be deployed on aircraft, balloons and satellites in the air. The scene where the network device and the terminal device are located is not limited in the embodiments of the present application.

[0093] Artificial intelligence (AI) technology can be combined with a wireless air interface to improve wireless network performance. For example, AI-based channel estimation and signal detection. Among them, signal detection is a process of extracting the received signal containing interference noise from the wireless channel; channel estimation is a process of estimating the model parameters of a hypothetical channel model from the received signal. For another example, AI-based end-to-end communication link design. For another example, an AI-based channel state information (CSI) feedback scheme, that is, encoding CSI through a neural network and feeding it back to the network device.

[0094] Taking the AI-based channel state information (CSI) feedback scheme as an example, the process of wireless communication between the network device and the terminal device using the AI model is introduced. As shown in (a) of Figure 2 The network device can deploy an AI encoder and a corresponding AI decoder. The network device can send or issue the AI encoder to the terminal device, or instruct the terminal device to obtain the AI encoder, so that the terminal device uses the AI encoder for CSI encoding in the future. As shown in (b) of Figure 2 The terminal device can use the AI encoder to encode the CSI and send the encoded information to the network device. Then, the network device can use the AI decoder to decode the information encoded by the terminal device to obtain the recovered information. It should be noted that the AI encoder can also be understood as an AI model for information encoding, and the AI decoder can also be understood as an AI model for information decoding.

[0095] AI models are various, and different AI models can be used in different application scenarios. Commonly, an AI model can be implemented based on a neural network model. The neural network model is a mathematical computational model that simulates the behavior characteristics of a human brain neural network and performs distributed parallel information processing. Some complex neural network models can contain a large amount of parameters or a large amount of calculations, and the capability (for example, computing capability, storage capability, or energy) of a terminal device can be limited. Therefore, before AI communication is performed between a network device and a terminal device, it is necessary to ensure that the AI capability of the terminal device can support execution (or referred to as running or processing) of an AI model sent by the network device, for example, the storage capability of the terminal device can accommodate the AI model, the computing capability of the terminal device can support the AI model to complete calculation within a required time, and the running power consumption (or referred to as energy consumption) of the terminal device for executing the AI model is within an expected acceptable range.

[0096] Generally, an upper limit t i of AI model calculation delay is known, and therefore, in a scheme, the matching of an AI model and the AI capability of a terminal device can be evaluated by comparing the computing capability C UE of the terminal device and the calculation complexity C M of the AI model. When C M / C UE +t th ≤ t i , it indicates that the AI model can complete calculation within a required delay, that is, the AI model complexity and the computing capability of the terminal device are matched, otherwise, it is considered that the AI model complexity and the computing capability of the terminal device are not matched. The computing capability C UE of the terminal device can be in units of floating-point operations per second (FLOPS), the calculation complexity C M of the AI model can be in units of floating-point operations (FLOP), and t th is a margin that can be configured in advance.

[0097] However, the internal structures of AI models for the same use may differ greatly, and in actual calculation, the calculation efficiency of AI models with different structures may differ greatly due to different hardware calculation rates, data scheduling delays, etc. In addition, due to the large amount of software conversion and optimization between AI models and hardware computing resources, different software conversion and optimization methods also bring different calculation efficiencies. In addition, different AI computing hardware implementations of terminal devices may also bring different AI model calculation efficiencies. Therefore, the above AI model structure design, software environment, hardware implementation, etc. may all cause hundreds of times of calculation efficiency deviation. Therefore, it is difficult to evaluate the actual matching of AI models and terminal device AI capabilities. th It may be difficult to evaluate the actual matching of AI models and terminal device AI capabilities. For example, t th Too large may waste computing resources, t th Too small may result in failure to complete execution of the AI model within the upper limit of the time delay; and t th The value may also be different.

[0098] In another solution, terminal hardware information, such as a simulator of a terminal device, can be configured on a matching server. After obtaining detailed information of an AI model and a calculation time delay, the matching server can obtain an evaluation result of the accurate matching of AI model complexity and terminal device AI capability. However, this solution needs to introduce a matching server and a corresponding interaction protocol, making the network structure and interaction process complex, and increasing the cost and evaluation time delay.

[0099] Therefore, the present application proposes an AI communication method and device, which can evaluate the matching of AI model complexity and device AI, and ensure the feasibility of using AI models for communication services.

[0100] It should be noted that the present application can be applied to a scenario in which any two devices (a first device and a second device) in a communication system communicate. The first device can have computing hardware that executes an AI model and is ready to obtain the AI model from the second device and use it, and the second device can have an AI model and be ready to send it to the second device for use. For example, the first device is a terminal device 121, a terminal device 122, or a terminal device 123 in Figure 1 , and the second device is a network device 111 in Figure 1 ; for another example, the first device is a terminal device 122 in Figure 1 , and the second device is a terminal device 125 in Figure 1 ; for yet another example, the first device is a network device 111 in Figure 1 , and the second device is a network device 112 in Figure 1 .

[0101] In the embodiments shown below, only for the convenience of understanding and description, taking the interaction between the network device and the terminal device as an example, the method provided by the embodiments of the present application is described in detail.

[0102] The method and related apparatus provided by the embodiments of the present application are described in detail below with reference to the accompanying drawings. It should be noted that the order of the embodiments of the present application only represents the order of the embodiments, and does not represent the advantages and disadvantages of the technical solutions provided by the embodiments.

[0103] Based on the above description, the AI communication method provided by the embodiments of the present application is suitable for Figure 1 The communication system is shown in the figure. As Figure 3 The method can include but is not limited to the following steps:

[0104] S301: The terminal device acquires AI capability information, which includes the time and / or energy consumption of the terminal device executing each reference AI unit in at least one reference AI unit.

[0105] For example, when the number of at least one reference AI unit is 3 (reference AI unit P1, reference AI unit P2, and reference AI unit P3), the AI capability information includes the time and / or energy consumption of the terminal device executing the reference AI unit P1, the time and / or energy consumption of the terminal device executing the reference AI unit P2, and the time and / or energy consumption of the terminal device executing the reference AI unit P3.

[0106] Specifically, the reference AI unit information can be standardized in advance or specified offline, and disclosed to the terminal device.

[0107] Wherein, the terminal device executing the reference AI unit can also be understood as the terminal device performing inference or training operation on the reference AI unit. For example, the reference AI unit can be at least one of a reference AI model, a reference AI module, and a reference AI operator. For example, the AI model can be a neural network model, the AI module can be a neural network module, and the AI operator can be a neural network operator. Common neural network models include, for example, ResNet series, MobileNet series, Transfbrmer series, etc. The neural network model includes a layer of operators (neural network operators), and the current neural network model (typically, such as ResNet series model) is often modularized design, that is, the neural network model includes a series of modules (neural network modules) and neural network operators. It can be considered that the AI module includes the AI operator, and the AI model includes the AI module and / or the AI operator.

[0108] Optionally, the time and / or energy consumption of the terminal device executing each reference unit can be represented by an actual value, an approximate value, a level (e.g., low / medium / high, or 1 / 2 / 3 / 4), or a range. Specifically, the at least one reference AI unit can include a first AI reference unit, and the time of the terminal device executing the first AI reference unit can be a first time value, a first time level, or a first time range, and the energy consumption of the terminal device executing the first AI reference unit can be a first energy consumption value, a first energy consumption level, or a first energy consumption range.

[0109] In an implementation manner, the terminal device can execute the at least one reference AI unit by itself and obtain the AI capability information thereof. In another implementation manner, the process of executing the at least one AI reference unit can be completed in a device with the same or similar AI capability as the terminal device or a simulator capable of truly embodying the AI capability of the terminal device. Specifically, the error of the calculation result obtained in the device with the same or similar AI capability as the terminal device or the simulator can not exceed an expected value. At this time, the terminal device can obtain the AI capability information from the other device or the simulator. In yet another implementation manner, the time and / or energy consumption information of the terminal device executing the at least one AI reference unit can be a factory setting.

[0110] S302: The terminal device sends the AI capability information to the network device. Correspondingly, the network device receives the AI capability information sent by the terminal device.

[0111] Optionally, the AI capability information sent by the terminal device can include the number of input data used by the terminal device in executing the at least one reference AI unit. The number of input data can also be referred to as the number of samples. Generally, the terminal device executes a batch of samples at a time when performing model training, and the sample batch size determines the number of samples for one training. When different numbers of input data are used, the terminal device will obtain different calculation efficiencies.

[0112] Optionally, the AI capability information sent by the terminal device can further include at least one of the following: the numerical precision of the input data used by the terminal device (or other devices with the same AI capability as the terminal device or a simulator) in executing each of the at least one reference AI unit; the numerical precision of the weight (or referred to as the weight, the coefficient) of each of the at least one reference AI unit; and the operation precision when the terminal device (or other devices with the same AI capability as the terminal device or a simulator) executes each of the at least one reference AI unit.

[0113] The neural network can be composed of neural units, and a neural unit can refer to an operation unit taking x s and an intercept 1 as input data, and the output of the operation unit can be: wherein s = 1, 2,..., n, n is a natural number, W s is the weight of x s , and b is the bias of the neural unit (which can also be regarded as a weight). The numerical precision of the input data and the weight values affects the computing efficiency of the terminal device. The numerical precision may be, for example, int (integer) 8, int 16, float (floating point) 16, and the like. The numerical precision can also be equivalently represented as the multiplication and addition operation precision, and the operation precision also affects the computing efficiency of the terminal device. The operation precision may be, for example, int 8, int 16, float 16, and the like

[0114] In one possible implementation, the resource used by each of the at least one reference AI unit executed by the terminal device is a part of all available computing resources of the terminal device. The running hardware of the terminal device can be of multiple types, for example, a central control unit (CPU), a microprocessor (GPU), an embedded neural network processor (NPU), or a field programmable gate array (FPGA), and the like. If the terminal device supports heterogeneous computing, the running hardware can also be a combination of the above-mentioned multiple types, wherein the heterogeneous computing refers to the distribution of the AI model on multiple types of computing units for execution. For example, the AI model is distributed on three types of computing units of CPU, GPU, and FPGA for execution. Therefore, the available computing resources of the terminal device can include one or more of CPU resources, GPU resources, and FPGA resources.

[0115] In another possible implementation, the resource used by each of the at least one reference AI unit executed by the terminal device is a part of all available computing resources of the terminal device. It can be understood that the proportion of the computing resources used by the terminal device to execute the reference AI unit affects the AI capability information (i.e., the time and / or energy consumption information of the terminal device for executing the at least one AI reference unit).

[0116] In this implementation, manner 1: the terminal device can convert the calculated completion time or consumed energy of executing each of the at least one reference AI unit according to the proportion of the used computing resources to all available computing resources, to obtain the finally reported AI capability information. Manner 2: the terminal device can report the proportion information of the used computing resources to all available computing resources to the network device, and then the network device can convert the AI capability information reported by the terminal device according to the resource proportion information. Optionally, the AI capability information sent by the terminal device can include (or indicate) the proportion of the computing resources used by the terminal device for executing each of the at least one reference AI unit to all available computing resources of the terminal device, or the terminal device can separately send the resource proportion information to the network device in other signaling.

[0117] Optionally, the process of the terminal device executing the at least one reference AI unit can be completed when the terminal device is manufactured.

[0118] It should be noted that the information included in the above AI capability information can be in the same signaling, or can be included in different signaling, which can be referred to as the proprietary capability information of the terminal device.

[0119] When the information included in the above AI capability information is in the same signaling, the terminal device can report the above AI capability information to the network device at one time, for example, the terminal device sends first information to the network device, the first information can include or indicate the above AI capability information, or the first information is the above AI capability information.

[0120] When the information included in the above AI capability information is in different signaling, the terminal device can report the AI capability information to the network device one or more times. Example one: the terminal device reports different capabilities such as the time and / or energy consumption of executing the AI unit, the number of input data, the precision of input data, etc. to the network device one or more times. For example, the terminal device sends first information to the network device, the first information includes or indicates the time and / or energy consumption of the terminal device for executing each of the at least one reference AI unit, the terminal device sends second information to the network device, the second information includes or indicates the number of input data used by each of the at least one reference AI unit, and the above AI capability information can include the first information and the second information. Example two: the terminal device reports the above AI capability information to the network device one or more times. For example, the terminal device sends first information to the network device, the first information includes or indicates the capability information such as the time and / or energy consumption of executing the AI unit, the number of input data, the precision of input data, etc., the terminal device sends second information to the network device, the second information includes or indicates the updated capability information, and the above AI capability information can include the first information and the second information.

[0121] In an embodiment, as shown in FIG. 3, after the step S302, i.e., after the terminal device sends the AI capability information to the network device, the method further comprises: Figure 3

[0122] S303: The network device sends configuration information to the terminal device. Correspondingly, the terminal device receives the configuration information sent by the network device. In an embodiment, the configuration information is response information of the AI capability information.

[0123] Specifically, the configuration information can satisfy one or more of the following: the configuration information can indicate that the terminal device starts the AI mode, and further, the AI communication between the terminal device and the network device can be performed. Alternatively, the configuration information can indicate at least one AI model, and further, the AI communication between the terminal device and the network device can be performed using the at least one AI model. Alternatively, the configuration information can indicate configuration parameters of the at least one AI model. Alternatively, the configuration information can indicate an acquisition method of the at least one AI model, for example, the configuration information can indicate a download address or an acquisition address of the at least one AI model. Wherein, the at least one AI model can be determined according to the AI capability information.

[0124] It can be understood that after the network device receives the AI capability information sent by the terminal device, the network device can also not respond.

[0125] In an embodiment, the terminal device sends the AI capability information to the network device, and the specific triggering method can include:

[0126] Method 1: The network device sends request information to the terminal device. Correspondingly, the terminal device receives the request information sent by the network device. Wherein, the request information is used to request the terminal device to send (or report) the AI capability information to the network device.

[0127] Method 2: The terminal device periodically sends the AI capability information to the network device. Alternatively, the terminal device sends the AI capability information to the network device according to a predefined time interval or a predefined specific time.

[0128] Method 3: When the terminal device accesses the network where the network device is located, the terminal device sends the AI capability information to the network device. Wherein, when the terminal device accesses the network where the network device is located, it can also be described as after the terminal device accesses the network where the network device is located, or within a certain time after the terminal device accesses the network where the network device is located. In addition, the terminal device accessing the network where the network device is located can also be understood as that the terminal device establishes a communication connection with the network device.

[0129] ​Option 4: When the computing resources available to the terminal device for executing the AI model change (e.g., increase or decrease), or the proportion of the computing resources available to the terminal device for executing the AI model changes, the terminal device sends AI capability information to the network device. As above, when the computing resources available to the terminal device for executing the AI model change, it can also be described as after the computing resources available to the terminal device for executing the AI model change, or within a certain time after the computing resources available to the terminal device for executing the AI model change, which will not be described here.

[0130] When the terminal device is in different application scenarios, the complexity of the operations to be completed is different, and the power / energy requirement, time requirement, and resource requirement that the terminal device can accept (or tolerate) for executing the AI model are also different. That is, in different application scenarios, the maximum time requirement, the maximum energy consumption requirement, or the resource usage requirement of the terminal device for executing the AI model are different, wherein the maximum time requirement can be how long the terminal device should take to execute the AI model, the maximum energy consumption requirement can be the maximum energy consumption allowed by the terminal device to consume after executing the AI model, and the resource usage requirement can be the maximum proportion of resources available to the terminal device that the terminal device is allowed to use for executing the AI model, or the hardware resource configuration that the terminal device can use for executing the AI model. It should be noted that the above-mentioned execution of the AI model by the terminal device refers to the process of executing the AI model by the terminal device in the AI communication mode, and does not limit the type of AI model executed by the terminal device.

[0131] Further, when the terminal device is in different application scenarios, the terminal device can send AI capability information to the network device, wherein the AI capability information can include at least one of a time upper limit value, an energy consumption upper limit value, and a resource usage situation for the terminal device for executing the AI model. It can be understood that the time upper limit value is also called time upper limit, maximum time, or maximum time requirement, the energy consumption upper limit value is also called energy consumption upper limit, maximum energy consumption, or maximum energy consumption requirement, and the resource usage situation can be, for example, a hardware resource configuration that can be used by the terminal device, a resource proportion upper limit value, etc. Among them, the resource proportion upper limit value is also called resource proportion upper limit, maximum resource proportion, or maximum resource proportion requirement. Optionally, before the terminal device sends the AI capability to the network device, the network device can also send query information to the terminal device, wherein the query information indicates that the terminal device reports one or more of the time upper limit value, the energy consumption upper limit value, and the resource usage situation for executing the AI model.

[0132] It should be understood that when the application scenario of the terminal device changes, the terminal device can report at least one of the time upper limit value, the energy consumption upper limit value, and the resource usage information of the terminal device for executing the AI model to the network device to inform the network device of at least one of the time budget requirement, the energy consumption limit, and the resource proportion consumption limit of the terminal device for executing the AI model. At this time, the time upper limit value information, the energy consumption upper limit value information, or the resource usage information and the information reported in step S302 can be carried in different signaling.

[0133] Through the above-described embodiments, the terminal device reports its AI capability information, so that the network device can evaluate the matching of the AI capability of the terminal device and the AI model (for example, the complexity of the AI model) through the AI capability information, and then determine whether to start the AI communication mode or issue a suitable AI model, which can improve the simplicity, efficiency, and accuracy of the evaluation result of the network device, and thus better perform AI communication.

[0134] In the above-described embodiments, after the terminal device sends its AI capability information to the network device, the network device can evaluate the AI capability of the terminal device and the complexity of the AI model, and determine the configuration information.

[0135] Figure 4 An example of a method flowchart for a network device to evaluate the matching degree of AI capability and AI model is given. As shown in Figure 4 Before the terminal device obtains the AI capability information, in step S401, the terminal device and the network device can obtain reference AI unit information.

[0136] As described above, the reference AI unit information can be standardized in advance or specified offline, and disclosed to the network device and the terminal device.

[0137] By way of example, an AI reference table can be standardized in advance or specified offline, where the AI reference table can include the names or numbers or indexes of N reference AI units (for example, AI models, AI modules, or AI operators), and the structure description parameters of the N reference AI units, where the structure description parameters of the N reference AI units can also be provided by referencing references or links. Through the AI reference table, the network device and the terminal device can uniformly understand the network structure of the N reference AI units. Optionally, the specific weight coefficient value (weight) of each AI unit can not be defined, or predefined, or a random number. The AI reference table is disclosed to the terminal device and the network device in advance. Specifically, in step S301, the at least one reference AI unit executed by the terminal device can be K reference AI units in the N reference AI units, where N is a positive integer, and K is a positive integer not greater than N.

[0138] S402: The network device acquires similarity information of M AI models. M is a positive integer.

[0139] Optionally, the M AI models can be pre-stored in the network device or other devices, or the M AI models are acquired by the network device from other devices, or the M AI models are converted or generated by the network device from existing AI models.

[0140] Specifically, the network device acquires M sets of similarity information of the M AI models and the K reference AI units.

[0141] In an implementation manner, the M AI models include a first AI model, and the K reference AI units include a first reference AI unit. The similarity information corresponding to the first AI model is a first set of similarity information in the M sets of similarity information, and the first set of similarity information includes first similarity information associated with a first proportion and / or a second proportion. The first proportion is a proportion of a total number of layers identical to the first reference AI unit in the first AI model to a total number of layers of the first AI model, and the second proportion is a proportion of a total amount of calculation of layers identical to the first reference AI unit in the first AI model to a total amount of calculation of the first AI model.

[0142] That is, the M sets of similarity information include M sets of similarity information corresponding to the M AI models. The first set of similarity information corresponding to the first AI model includes K sets of similarity information corresponding to the first AI model and the K reference AI units.

[0143] Table 1 shows a common neural network model, a residual neural network (ResNet) model. An AI model can include an AI module. For example, the ResNet-34 model shown in Table 1 includes AI modules Further, an AI module can include an AI operator. For example, the AI module includes an AI operator [3x3, 64] (a convolution operation operator).

[0144] Table 1: ResNet neural network model

[0145]

[0146] For example, taking the first AI model as the ResNet-50 model (i.e., the ResNet model with 50 layers) as an example, when the first reference AI unit is a model, for example, the ResNet-34 model shown in Table 1, since the same layers in the ResNet-50 model and the ResNet-34 model are three layers [3x3, 64], four layers [3x3, 128], six layers [3x3, 256], three layers [3x3, 512], and [7x7, 64, stride2], [3x3 max pool, stride 2], a total of 18 layers, and the proportion of the same number of layers in the total number of layers of the ResNet-50 model is 18 / 50 = 9 / 25, i.e., the first proportion is 9 / 25. When the first reference AI unit is a module, for example, the module When the first reference AI unit is an operator, for example, the operator [3x3, 256], the same layer in the ResNet-50 model and the operator is [3x3, 256], a total of 6 layers, and the proportion of the same layer is 6 / 50 = 3 / 25, i.e., the first proportion is 3 / 25. Similarly, the total calculation amount (in FLOP) of these same layers is divided by the total calculation amount 3.8x10 9 FLOPs of the ResNet-50 model, i.e., the second proportion is obtained.

[0147] In another implementation manner, the M AI models include a first AI model, and the K reference AI units include a first reference AI unit. The similarity information corresponding to the first AI model is a first group of similarity information in the M groups of similarity information, and the first group of similarity information includes first similarity information, and the first similarity information is associated with a first proportion and / or a second proportion, where the first proportion is a proportion of a total number of units same as the first reference AI unit in the first AI model in a total number of units of the first AI model, and the second proportion is a proportion of a total calculation amount of units same as the first reference AI unit in the first AI model in a total calculation amount of the first AI model.

[0148] For example, taking the first AI model as the ResNet-50 model shown in Table 1 as an example, when the first reference AI unit is a model, for example, the ResNet-34 model shown in Table 1, since the ResNet-50 model is different from the ResNet-34 model, i.e., there is no unit same as the first reference AI unit in the first AI model, at this time, the first proportion is 0. When the first reference AI unit is a module, for example, the module When the ResNet-50 model includes 3 of the module, the first ratio is 3x3 / 50 = 9 / 50. When the first reference AI unit is the module, for example, the module When the ResNet-50 model includes 0 of the module, the first ratio is 0. When the first reference AI unit is the operator, for example, the operator [3x3, 64], the ResNet-50 model includes 3 of the operator, and the first ratio is 3 / 50.

[0149] Optionally, the network device can perform local calculation to obtain the similarity information of the M AI models. Alternatively, the calculation process can also be completed in other devices, and then the network device can obtain the calculation result from the other devices.

[0150] S403: The network device performs AI capability and AI model matching degree evaluation.

[0151] After the foregoing step S302, that is, after the network device receives the AI capability information sent by the terminal device, the network device can evaluate the matching between the AI capability of the terminal device and the M AI models based on the AI capability information received in step S302.

[0152] In some embodiments, the network device can obtain one or more of the expected completion time, the expected power consumption, or the expected resource proportion used by the M models in the terminal device through analogy evaluation. For example, when the similarity is represented by the foregoing first ratio, it is assumed that the total number of layers of the first AI model is L0, and in step S402, it is assumed that the network device obtains the similarity S1 between the first AI model and the reference AI model K1, the similarity S2 between the first AI model and the reference AI module K2, and the similarity S3 between the first AI model and the reference AI operator K3. In step S401, the terminal device and the network device can learn that the total number of layers of K1, K2, and K3 is L1, L2, and L3 (L3 = 1, the number of layers of the operator is 1) respectively. In step S302, the AI capability information sent by the terminal device to the network device includes the completion time t1, t2, and t3 of the terminal device executing the foregoing reference AI model, reference AI module, and reference AI operator respectively. Then, the network device can evaluate the expected completion time of the terminal device executing the first AI model as When the similarity is represented by the foregoing second ratio, L0, L1, L2, and L3 can represent the total calculation amount, and the network device can also obtain the foregoing result by using a similar scheme, which is not described herein again. The network device can also obtain the expected energy consumption and the expected resource proportion used by the terminal device executing the first AI model by using analogy evaluation, which is not described herein again.

[0153] It should be understood that the above description is an exemplary method of the network device for evaluating the matching degree of the terminal device AI capability and the AI model. In another embodiment, the network device can perform more detailed analysis and obtain more accurate evaluation results by using the M models and the detailed structure of the reference AI unit, and the present application does not limit this.

[0154] Based on one or more of the expected completion time, the expected power consumption, or the expected proportion of resources obtained by the above evaluation, and one or more of the time upper limit value information, the energy consumption upper limit value information, or the resource usage information sent by the terminal device for executing the AI model, the network device can determine whether the M models match the AI capability of the terminal device. For example, if the expected completion time of the first AI model in the M models does not exceed the time budget requirement of the terminal device, and the expected energy consumption does not exceed the energy consumption limit of the terminal device, and the expected proportion of resources consumed does not exceed the resource proportion limit of the terminal device, it can be determined that the first AI model matches the AI capability of the terminal device.

[0155] Further, as described in the foregoing step S303, the network device can send configuration information to the terminal device.

[0156] In step S303, when the configuration information indicates at least one AI model, or the configuration information indicates the configuration parameters of at least one AI model, or the configuration information indicates the obtaining method of at least one AI model, the network device determines at least one AI model according to the AI capability information before sending the configuration information to the terminal device. Specifically, in one case, the network device has only one AI model facing the current application, and if the AI model matches the AI capability of the terminal device, the AI model is the AI model determined by the network device. In another case, the network device has more than one AI model facing the current application, and the network device can select at least one AI model from the AI models matching the AI capability of the terminal device to send or configure to the terminal device, wherein the selection principle of the network device can be the shortest expected completion time, the least expected energy consumption, the smallest expected proportion of resources used, or a random principle. It can be understood that if the network device can determine at least one AI model, it is considered that the network device can start the AI mode.

[0157] In step S303, when the configuration information indicates the obtaining method of at least one AI model, specifically, the network device can indicate the download address or obtaining address of at least one AI model. As described above, the M AI models can be pre-stored in other devices (for example: a third device), at this time, the network device can instruct the terminal device to obtain the at least one AI model from the third device.

[0158] It should be noted that, Figure 4In the embodiments shown, the specific implementation process of steps S301, S302 and S303 can be referred to Figure 3 The related expressions in the embodiments shown are not described here. In the above embodiments, the terminal device sends the data obtained by executing the reference AI model, AI module or AI operator to the network device as AI capability information, which can accurately represent the AI capability of the terminal device, so that the network device obtains more accurate evaluation results, and then decides whether to start AI communication or issues an AI model.

[0159] The following will briefly describe the embodiments of the present application taking the AI-based CSI feedback scheme as an example.

[0160] The network device and the terminal device can first obtain a reference AI unit. Based on the reference AI unit, the terminal device can obtain AI capability information as described in step S301, and the network device can obtain AI model similarity information as described in step S402. Then, the terminal device can send the AI capability information it obtains to the network device, so that the network device can perform AI capability and AI model matching degree evaluation. After the evaluation is completed, the network device can send a suitable AI encoder to the terminal device. After the terminal device receives the AI encoder, it can use the AI encoder to encode the CSI and send the encoded information to the network device. Then, the network device can use the AI decoder corresponding to the AI encoder to decode the information encoded by the terminal device, to obtain the restored information.

[0161] The embodiments of the present application also provide an AI communication method, which is suitable for Figure 1 The communication system shown. As Figure 5 The method can include but is not limited to the following steps:

[0162] S501: The network device sends AI model information to the terminal device, and correspondingly, the terminal device receives the AI model information sent by the network device. The AI model information can indicate the complexity information of each AI model in M AI models.

[0163] Exemplarily, the complexity information of each AI model in the M AI models can be represented by similarity data. For example, the AI model information indicates M sets of similarity information (or similarity data) corresponding to the M AI models. Each set of similarity information in the M sets of similarity information is the similarity information between one AI model in the M AI models and K reference AI units, wherein the K reference AI units belong to N reference AI units, M and N are positive integers, and K is a positive integer less than or equal to N.

[0164] For example, when M=3 and K=5, the AI model information includes 3 sets of similarity information (1st set of similarity information, 2nd set of similarity information, and 3rd set of similarity information) corresponding to 3 AI models (Q1, Q2, and Q3), where the 1st set of similarity information is the similarity information between the AI model Q1 and the 5 reference AI units, the 2nd set of similarity information is the similarity information between the AI model Q2 and the 5 reference AI units, and the 3rd set of similarity information is the similarity information between the AI model Q3 and the 5 reference AI units.

[0165] Optionally, the M AI models can be pre-stored in the network device or other devices, or the M AI models are obtained by the network device from other devices, or the M AI models are converted or generated by the network device from existing AI models.

[0166] For example, the N reference AI units can be in an AI reference table, and the specific content of the AI reference table can be referred to the above description.

[0167] In an implementation manner, the M AI models include a first AI model, and the K reference AI units include a first reference AI unit. The similarity information corresponding to the first AI model is the 1st set of similarity information in the M sets of similarity information, and the 1st set of similarity information includes first similarity information associated with a first proportion and / or a second proportion, where the first proportion is a proportion of a total number of layers in the first AI model that are the same as the first reference AI unit in a total number of layers of the first AI model, and the second proportion is a proportion of a total amount of calculation of layers in the first AI model that are the same as the first reference AI unit in a total amount of calculation of the first AI model.

[0168] In another implementation manner, the M AI models include a first AI model, and the K reference AI units include a first reference AI unit. The similarity information corresponding to the first AI model is the 1st set of similarity information in the M sets of similarity information, and the 1st set of similarity information includes first similarity information associated with a first proportion and / or a second proportion, where the first proportion is a proportion of a total number of layers in the first AI model that are the same as the first reference AI unit in a total number of layers of the first AI model, and the second proportion is a proportion of a total amount of calculation of layers in the first AI model that are the same as the first reference AI unit in a total amount of calculation of the first AI model. For specific examples of the similarity information, refer to the above description, which will not be described here.

[0169] Optionally, the M sets of similarity information corresponding to the M AI models can be represented by actual numerical values, approximate numerical values, levels, or ranges. For example, the first similarity information described above can be a first similarity numerical value, a first similarity level, or a first similarity range.

[0170] In an implementation manner, the network device can locally calculate the M sets of similarity information. In another implementation manner, the process of calculating the M sets of similarity information can be completed at a third device (e.g., another network device or a third-party server), and the network device can obtain the M sets of similarity information from the third device.

[0171] Optionally, the AI model information sent by the network device can include at least one of the following: a numerical precision of input data used by the network device or the third device when executing each of the M AI models; a numerical precision of a weight (or a coefficient) of each of the M AI models; and an operation precision when the network device or the third device executes each of the M AI models.

[0172] Optionally, the AI model information sent by the network device can further include a total number of layers and a total amount of calculation of each of the M AI models. It can be understood that, by obtaining more detailed information of the M AI models, the terminal device can more accurately evaluate the complexity of the M AI models.

[0173] The time budget information for executing the AI model is different in different application scenarios. Further, the AI model information described above can further include an upper limit value of time for executing the AI model, that is, the time budget information for executing the AI model to be delivered, that is, the time required to execute the AI model. It can be understood that the upper limit value of time is also referred to as the upper limit of time, the maximum time, or the maximum time requirement.

[0174] It should be noted that the information included in the AI model information described above can be in the same signaling, or can be included in different information. When the information included in the AI model information described above is in the same signaling, the network device can send the AI model information described above to the terminal device at one time. When the information included in the AI model information described above is in different signaling, the network device can send the AI model information described above to the terminal device one or more times.

[0175] S502: The terminal device sends feedback information to the network device. Correspondingly, the network device receives the feedback information sent by the terminal device.

[0176] Exemplarily, the feedback information can satisfy one or more of the following: the feedback information can be used to request to start the AI communication mode. Alternatively, the feedback information can include or indicate the evaluation result of at least one AI model in the M AI models. Alternatively, the feedback information can request the network device to send at least one AI model to the terminal device, wherein the at least one AI model belongs to the M AI models.

[0177] In an implementation, as shown in FIG. 5A, before the foregoing step S501, that is, before the network device sends the AI model information to the terminal device, the method further includes: Figure 5

[0178] S500: The network device obtains the AI model information.

[0179] As described above, the network device can locally calculate the M sets of similarity information to obtain the AI model information. Alternatively, the network device can obtain the M sets of similarity information from a third device.

[0180] In an implementation, as shown in FIG. 5B, after the foregoing step S502, that is, after the terminal device sends the feedback information to the network device, the method further includes: Figure 5

[0181] S503: The network device sends configuration information to the terminal device. Correspondingly, the terminal device receives the configuration information sent by the network device. The configuration information is response information of the feedback information.

[0182] Specifically, the configuration information can satisfy one or more of the following: the configuration information can indicate the terminal device to start the AI mode, and further, the AI communication between the terminal device and the network device can be performed. Alternatively, the configuration information can indicate at least one AI model, and further, the AI communication between the terminal device and the network device can be performed using the at least one AI model. Alternatively, the configuration information can indicate the obtaining method of the at least one AI model, for example, the configuration information can indicate the download address or obtaining address of the at least one AI model. The at least one AI model is determined according to the feedback information.

[0183] When the configuration information indicates the obtaining method of the at least one AI model, as described above, the M AI models can be pre-stored in other devices (for example: a third device), and at this time, the network device can instruct the terminal device to obtain the at least one AI model from the third device.

[0184] It can be understood that after the network device receives the feedback information sent by the terminal device, the network device can also not respond.

[0185] ​​In an implementation, before the foregoing step S501, that is, before the network device sends the AI model information to the terminal device, the method further includes: the terminal device sends request information to the network device, the request information being used to request the network device to send the AI model information to the terminal device.

[0186] In the foregoing embodiments, after the network device sends the M sets of similarity information of the M AI models to the terminal device, the terminal device can perform complexity evaluation on the M AI models, and further determine the matching condition of the M AI models and the AI capability of the terminal device.

[0187] In the foregoing embodiments, before the foregoing step S502, that is, before the terminal device sends the feedback information to the network device, the method further includes: the terminal device determines the feedback information according to the AI model information received in the step S501 and the AI capability information of the terminal device. The following will exemplarily introduce a method for the terminal device to determine the feedback information through specific embodiments.

[0188] Figure 6 An exemplary method flowchart for the terminal device to perform AI capability and AI model matching degree evaluation is given. As shown in Figure 6 Before the network device acquires the AI model information, in step S601, the network device and the terminal device can acquire reference AI unit information.

[0189] S602: After the terminal device acquires the reference AI unit information, the terminal device acquires AI capability information.

[0190] The specific implementation of the network device and the terminal device acquiring the reference AI unit information in the step S601 can refer to the description in the foregoing embodiment step S401, and the specific implementation of the terminal device acquiring the AI capability information in the step S602 can refer to the description in the foregoing embodiment step S301, which will not be repeated here.

[0191] S603: The terminal device performs AI capability and model matching degree evaluation.

[0192] After the foregoing step S501, that is, after the terminal device receives the AI model information sent by the network device, the terminal device can evaluate the matching condition of the AI capability of the terminal device and the M AI models based on the AI model information received in the step S501.

[0193] Similar to the foregoing embodiments, the terminal device can also obtain one or more of the expected completion time, the expected power consumption, or the expected proportion of resources used in the terminal device by executing the M models through analog evaluation. For example, when the similarity is expressed by the first proportion described above, in step S501, the network device sends the AI model information to the terminal device, including the total number of layers L0 of the first AI model, and the similarity information: the similarity of the first AI model with the reference AI model K1 is S1, the similarity of the first AI model with the reference AI module K2 is S2, and the similarity of the first AI model with the reference AI operator K3 is S3. In step S601, the terminal device and the network device can know that the total number of layers of K1, K2, and K3 is L1, L2, and L3 (L3 = 1, the number of layers of the operator is 1) respectively. In step S602, the terminal device obtains the completion time of executing the above-mentioned reference AI model, reference AI module, and reference AI operator as t1, t2, and t3 respectively. Then the terminal device can evaluate the expected completion time of executing the first AI model as When the similarity is expressed by the second proportion described above, L0, L1, L2, and L3 can represent the total amount of calculation, and the network device can also use a similar scheme to obtain the above-mentioned results, which will not be described here. The method for the terminal device to evaluate the expected energy consumption and the expected proportion of resources used in executing the first AI model can also be obtained by analog evaluation, which will not be described here.

[0194] Based on one or more of the expected completion time, the expected power consumption, or the expected proportion of resources used obtained through the above-mentioned evaluation, and the time limit value (i.e. the time budget information) for executing the AI model sent by the network device, the terminal device can determine whether the M models match its AI capability. For example, if the expected completion time of the first AI model in the M models does not exceed the time budget requirement of the terminal device, and the expected energy consumption does not exceed the energy consumption limit of the terminal device, and the expected proportion of resources consumed does not exceed the resource proportion limit of the terminal device, the terminal device can determine that the first AI model matches the AI capability of the terminal device. Wherein, the time budget requirement of the terminal device can be the time budget information for the terminal device to execute the AI model sent by the network device, or the local time budget requirement of the terminal device; the energy consumption limit of the terminal device can be the local energy consumption limit of the terminal device; and the resource proportion limit of the terminal device can be the local resource proportion limit of the terminal device.

[0195] Further, the terminal device can send feedback information to the network device. For example, in step S502, the feedback information can indicate at least one AI model. Specifically, the terminal device can select at least one AI model from the models matching its AI capability, and apply the network device to send the model through the feedback information. The selection principle of the terminal device can be the shortest expected completion time, the least expected energy consumption, the smallest expected resource proportion, or a random principle.

[0196] It should be noted that, Figure 6 In the embodiments shown, the specific implementation process of steps S500, S501, S502, and S503 can be referred to Figure 5 In the embodiments shown, the specific implementation process of steps S500, S501, S502, and S503 can be referred to

[0197] In the above embodiments, after the network device transmits the complexity of the AI model to be delivered to the terminal device, the terminal device can more simply, efficiently, and accurately evaluate the matching of the AI model to be delivered and the AI capability of the terminal device, and then determine whether to request to start the AI model or request to deliver the AI model, which can improve the accuracy and efficiency of the terminal device in evaluating the matching degree of the AI capability and the AI model.

[0198] Through the method provided by the embodiments of the present application, the terminal device can send its AI capability information to the network device, so that the network device can accurately and efficiently evaluate the matching of the AI model and the AI capability of the terminal device, or the network device can send the AI model information to the terminal device, so that the terminal device can accurately and efficiently evaluate the matching of the AI model and its AI capability. In addition, the method provided by the embodiments of the present application does not need to introduce additional server devices and interaction protocols, and can realize the evaluation of the matching degree of the AI capability and the AI model between the terminal device and the network device.

[0199] It can be understood that, in order to implement the functions in the above embodiments, the network device and the terminal device include hardware structures and / or software modules corresponding to each function. Those skilled in the art should easily realize that, in combination with the units and method steps of the examples described in the embodiments disclosed in the present application, the present application can be realized in the form of hardware or a combination of hardware and computer software. Whether a certain function is executed in hardware or computer software driven hardware depends on the specific application scenario and design constraints of the technical solution.

[0200] Figure 7 and Figure 8The diagram illustrates the possible structures of communication devices provided in the embodiments of this application. These communication devices can be used to implement the functions of the terminal device or network device in the above method embodiments, and thus can also achieve the beneficial effects of the above method embodiments. In the embodiments of this application, the communication device can be a terminal device or a network device, or it can be a module (such as a chip) applied to the terminal device or network device.

[0201] like Figure 7 As shown, the communication device 700 includes a transceiver unit 701 and a processing unit 702. The processing unit 702 is used to invoke the transceiver unit 701 to receive information from other communication devices or to send information to other communication devices. The transceiver unit 701 may further include a receiving unit and a sending unit; the receiving unit is used to receive information from other communication devices, and the sending unit is used to send information to other communication devices. The communication device 700 is used to implement the above-described... Figure 3 , Figure 4 , Figure 5 , Figure 6 The methods illustrated in this embodiment demonstrate the functions of the terminal device or network device. Figure 4 The embodiments are based on Figure 3 In the example, Figure 6 The embodiments are based on Figure 5 Examples of embodiments. The following are examples. Figure 3 Examples and Figure 5 The embodiments are provided as examples to illustrate the operations performed by the transceiver unit 701 and the processing unit 702 respectively. The operations performed by the two units in other embodiments can be obtained by referring to the method embodiments.

[0202] When the communication device 700 is used to achieve Figure 3 In the method embodiment shown, the terminal device functions as follows: a transceiver unit 701 is used to send AI capability information to the network device, the AI ​​capability information including the time and / or energy consumption of the terminal device executing each of at least one reference AI unit; a processing unit 702 is used to acquire the AI ​​capability information. When the communication device 700 is used to implement... Figure 3 In the method embodiment shown, the network device functions as follows: a transceiver unit 701 is used to receive AI capability information sent by a terminal device, the AI ​​capability information including the time and / or energy consumption of the terminal device executing each of at least one reference AI unit; the transceiver unit 701 is also used to send configuration information to the terminal device, wherein the configuration information instructs the terminal device to start an AI mode, or the configuration information instructs at least one AI model, or the configuration information instructs configuration parameters of at least one AI model, or the configuration information instructs a method for obtaining at least one AI model, wherein at least one AI model is determined based on the AI ​​capability information.

[0203] When the communication apparatus 700 is configured to implement the functions of the terminal device in the method embodiment shown in Figure 5 , the transceiver 701 is configured to receive AI model information sent by the network device, the AI model information including M sets of similarity information corresponding to M AI models, each set of similarity information in the M sets of similarity information being similarity information of one of the M AI models and K reference AI units, where M and K are positive integers; and the transceiver 701 is further configured to send feedback information to the network device according to the AI model information. When the communication apparatus 700 is configured to implement the functions of the network device in the method embodiment shown in Figure 5 , the transceiver 701 is configured to send AI model information to the terminal device, the AI model information including M sets of similarity information corresponding to M AI models, each set of similarity information in the M sets of similarity information being similarity information of one of the M AI models and K reference AI units, where M and K are positive integers; and the transceiver 701 is further configured to receive feedback information sent by the terminal device, where the feedback information is determined according to the AI model information.

[0204] For more detailed descriptions of the transceiver 701 and the processing unit 702 described above, refer directly to the relevant descriptions in the method embodiments shown in Figure 3 , Figure 5 , which will not be repeated here.

[0205] Based on the same technical concept, as shown in Figure 8 , the embodiments of the present application further provide a communication apparatus 800. The communication apparatus 800 includes an interface circuit 801 and a processor 802. The interface circuit 801 and the processor 802 are coupled to each other. It can be understood that the interface circuit 801 can be a transceiver or an input / output interface. Optionally, the communication apparatus 800 can further include a memory 803 for storing instructions executed by the processor 802, or storing input data required by the processor 802 to run instructions, or storing data generated after the processor 802 runs instructions.

[0206] When the communication apparatus 800 is configured to implement the method shown in Figure 3 , Figure 4 , Figure 5 , Figure 6 , the processor 802 is configured to implement the functions of the processing unit 702 described above, and the interface circuit 801 is configured to implement the functions of the transceiver 701 described above.

[0207] When the communication device is a chip applied to a terminal device, the terminal device chip implements the functions of the terminal device in the method embodiments. The terminal device chip receives information from other modules (such as a radio frequency module or an antenna) in the terminal device, and the information is sent by the network device to the terminal device. Alternatively, the terminal device chip sends information to other modules (such as a radio frequency module or an antenna) in the terminal device, and the information is sent by the terminal device to the network device.

[0208] It can be understood that the processor in the embodiments of the present application can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field programmable gate arrays (FPGAs) or other programmable logic devices, transistor logic devices, hardware components or any combination thereof. The general-purpose processor can be a microprocessor or any conventional processor.

[0209] Based on the above embodiments, the embodiments of the present application provide a communication system, which can include terminal devices and network devices and the like related to the above embodiments.

[0210] The embodiments of the present application also provide a computer readable storage medium for storing a computer program, which, when executed by a computer, can implement the downlink scheduling method provided by the above method embodiments.

[0211] The embodiments of the present application also provide a computer program product for storing a computer program, which, when executed by a computer, can implement the downlink scheduling method provided by the above method embodiments.

[0212] The embodiments of the present application also provide a chip, which includes a processor coupled with a memory, and is used to call a program in the memory to make the chip implement the downlink scheduling method provided by the above method embodiments.

[0213] The embodiments of the present application also provide a chip, which is coupled with a memory, and is used to implement the downlink scheduling method provided by the above method embodiments.

[0214] The method steps in the embodiments of the present application can be implemented by hardware or by a processor executing software instructions. The software instructions can be composed of corresponding software modules, which can be stored in a random access memory (RAM), a flash memory, a read-only memory (ROM), a programmable ROM (PROM), an erasable PROM (EPROM), an electrically EPROM (EEPROM), a register, a hard disk, a mobile hard disk, a CD-ROM, or any other form of storage medium well known in the art. An exemplary storage medium is coupled to a processor, so that the processor can read information from, and write information to, the storage medium. Of course, the storage medium can also be a component of the processor. The processor and the storage medium can be located in an ASIC. In addition, the ASIC can be located in a network device or a terminal device. Of course, the processor and the storage medium can also exist as discrete components in a network device or a terminal device.

[0215] In the above embodiments, all or part of the embodiments can be implemented by software, hardware, firmware, or any combination thereof. When implemented by software, all or part of the embodiments can be implemented in the form of a computer program product. The computer program product includes one or more computer programs or instructions. When the computer programs or instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of the present application are performed. The computer can be a general purpose computer, a special purpose computer, a computer network, a network device, a user equipment, or other programmable apparatus. The computer programs or instructions can be stored in a computer readable storage medium or transferred from one computer readable storage medium to another computer readable storage medium, for example, the computer programs or instructions can be transferred from one website site, computer, first device, or data center to another website site, computer, first device, or data center through wired or wireless manner. The computer readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a first device, data center, etc. that integrates one or more available media. The available media can be a magnetic medium, such as a floppy disk, a hard disk, a magnetic tape; an optical medium, such as a digital video disc (DVD); or a semiconductor medium, such as a solid state drive (SSD).

[0216] It is apparent that a person skilled in the art can make various changes and modifications to the embodiments of the application without departing from the scope of the application. Therefore, if these modifications and changes of the embodiments of the application belong to the scope of the claims of the application and their equivalent technologies, the application also intends to include these modifications and changes.

Claims

1. An artificial intelligence (AI) communication method comprising: The method comprises: The first device acquires AI capability information, the AI capability information comprising time and / or energy consumption of each of at least one predefined standardized modular reference AI unit executed by the first device, the reference AI unit being at least one of a reference AI model, a reference AI module, and a reference AI operator; The first device sends the AI capability information to a second device, the AI capability information further comprising at least one of: floating-point or integer numerical precision of input data used by each of the at least one reference AI unit executed by the first device; floating-point or integer numerical precision of weights of each of the at least one reference AI unit executed by the first device; operational precision of each of the at least one reference AI unit executed by the first device, the operational precision comprising precision of multiplication and addition operations.

2. The method of claim 1, wherein, After the first device sends the AI capability information to the second device, the method further comprises: The first device receives configuration information sent by the second device, wherein the configuration information indicates that the first device starts an AI communication mode, or the configuration information indicates at least one AI model, or the configuration information indicates configuration parameters of at least one AI model, or the configuration information indicates an acquisition method of at least one AI model.

3. The method according to claim 1 or 2, characterized in that, Before the first device sends the AI capability information to the second device, the method further comprises: The first device receives request information from the second device, the request information being used to request the first device to send the AI capability information to the second device.

4. The method according to any one of claims 1 to 3, characterized in that, The first device sends the AI capability information to the second device, comprising: The first device periodically sends the AI capability information to the second device; or When the first device accesses a network in which the second device is located, the first device sends the AI capability information to the second device; or When the first device establishes a communication connection with the second device, the first device sends the AI capability information to the second device; or When a computing resource used by the first device to execute an AI model changes, the first device sends the AI capability information to the second device.

5. An artificial intelligence (AI) communication method, comprising: The method comprises: The second device receives AI capability information sent by the first device, the AI capability information comprising time and / or energy consumption of each of at least one predefined standardized modular reference AI unit executed by the first device, the reference AI unit being at least one of a reference AI model, a reference AI module, and a reference AI operator, the AI capability information further comprising at least one of: floating-point or integer numerical precision of input data used by each of the at least one reference AI unit executed by the first device; floating-point or integer numerical precision of weights of each of the at least one reference AI unit executed by the first device; The first device performs an operation precision of each of the at least one reference AI unit, the operation precision including precision of multiplication and addition operations.

6. The method of claim 5, wherein, After the second device receives the AI capability information sent by the first device, the method further includes: The second device sends configuration information to the first device, wherein the configuration information indicates that the first device starts an AI communication mode, or the configuration information indicates at least one AI model, or the configuration information indicates configuration parameters of at least one AI model, or the configuration information indicates an acquisition method of at least one AI model.

7. The method of claim 6, wherein, When the configuration information indicates at least one AI model, or the configuration information indicates configuration parameters of at least one AI model, or the configuration information indicates an acquisition method of at least one AI model, before the second device sends the configuration information to the first device, the method further includes: The second device determines the at least one AI model according to the AI capability information.

8. The method according to any one of claims 5 to 7, characterized in that, Before the second device receives the AI capability information sent by the first device, the method further includes: The second device sends request information to the first device, the request information being used to request the first device to send the AI capability information to the second device.

9. The method of claim 2 or 6, wherein, The configuration information is response information of the AI capability information.

10. The method according to any one of claims 1 to 9, characterized in that, The time for the first device to execute a first AI reference unit in the at least one reference AI unit is a first time value, a first time level or a first time range, and the energy consumption of the first device for executing the first AI reference unit is a first energy consumption value, a first energy consumption level or a first energy consumption range.

11. The method according to any one of claims 1 to 10, characterized in that, The AI capability information includes one or more of the following: The first device executes each of the at least one reference AI unit using a number of input data; The first device uses an upper limit value of time for executing an AI model; The first device uses an upper limit value of energy consumption for executing an AI model; The first device uses a resource usage for executing an AI model.

12. The method according to any one of claims 1 to 11, characterized in that, The resource used by the first device for executing each of the at least one reference AI unit is all available computing resources of the first device.

13. A first device, comprising: Comprise: One or more processors and one or more memories; The one or more memories are coupled to the one or more processors, and the one or more memories are configured to store computer program codes, the computer program codes comprising computer instructions, when the one or more processors execute the computer instructions, causing the first device to execute the method of any one of claims 1-4 or claims 9-12.

14. A second device, comprising: Comprise: One or more processors and one or more memories; The one or more memories are coupled to the one or more processors, and the one or more memories are configured to store computer program codes, the computer program codes comprising computer instructions, when the one or more processors execute the computer instructions, causing the second device to execute the method of any one of claims 5-12.

15. A computer readable storage medium characterized by: The computer readable storage medium stores computer executable instructions which, when invoked by the computer, cause the computer to perform the method of any of claims 1-12.

16. A computer program product comprising instructions, characterized in that, The computer program product, when run on a computer, causes the computer to perform the method of any of claims 1-12.

17. A chip, characterized by The chip is coupled with a memory for reading and executing program instructions stored in the memory to implement the method of any of claims 1-12.

Images