Dsr transmission method, device, communication system and storage medium

Abstract

The application provides a DSR transmission method, device, communication system and storage medium, which can be applied to the technical field of wireless communication. The scheme defines a new DSR triggering condition. When the terminal device updates the sending window, the terminal device calculates the remaining bytes and the remaining time length of the current task in real time. Once the remaining time length is lower than the triggering threshold corresponding to the service type, the terminal device immediately triggers the DSR and reports the remaining bytes and the remaining time length of the task to the network device, and different types of non-video traffic services correspond to different triggering thresholds. The network device reserves resources in advance for the remaining bytes of the task according to the information. When the data of the service arrives, it can directly fall on the reserved resources, without the need to trigger the scheduling request again, solving the queuing delay problem and improving the scheduling efficiency and the transmission reliability.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of wireless communication, and in particular to a DSR transmission method, device, communication system and storage medium. BACKGROUND

[0002] With the rapid development of artificial intelligence (AI) technology, various large language models (LLM), AI-driven image processing tools and automatic driving applications are widely used. These AI applications not only provide services for AI video services, but also provide services for AI non-video services such as voice question and answer, image upload and text interaction. The total delay of the AI process needs to be limited within a very short time, and the AI inference of the cloud device usually consumes a long time, leaving only a small delay budget for network communication between the terminal device and the cloud device.

[0003] In the extended reality (XR) enhancement mechanism, the delay status report (DSR) based on the packet data convergence protocol (PDCP) discard timer is designed for AI video services. AI video services are transmitted based on a protocol without congestion control (such as UDP), but AI non-video services are transmitted based on a reliable transmission protocol with congestion control (such as TCP or QUIC), and if the DSR mechanism is directly applied to non-video services, it will cause transmission delay. SUMMARY

[0004] The present application provides a DSR transmission method, device, communication system and storage medium, which is used to solve the problem that the traditional DSR transmission mechanism is not compatible with mobile AI non-video.

[0005] To achieve the above-mentioned purpose, the present application adopts the following technical solutions:

[0006] In a first aspect, a DSR transmission method is provided. The method can be applied to a terminal device. The method can be executed by the terminal device, or can also be executed by a component (such as a circuit, a chip or a chip system, etc.) configured in the terminal device, and can also be implemented by a logic module or software capable of implementing all or part of the functions of the terminal device.

[0007] The method can include:

[0008] The data packet of the first AI non-video task is obtained, the first AI non-video task being an AI non-video task generated based on discrete events or instructions and relying on AI technology to complete inference on a cloud device. The data packet of the first AI non-video task is started to be sent to the cloud device through the network device. In a case where a remaining time length of the first AI non-video task is less than or equal to a first trigger time threshold, a DSR is sent to the network device, the DSR can include a remaining data amount (such as a remaining byte), the remaining data amount being an estimated data amount of the data packet to be transmitted in the first AI non-video task, and the remaining data amount being used to trigger the network device to reserve resources. A first uplink grant is received from the network device. On resources indicated by the first uplink grant, the data packet to be transmitted in the first AI non-video task is continued to be sent to the cloud device through the network device.

[0009] In the above scheme, the DSR trigger condition is changed from "PDCP discard timer timeout" to "task remaining time length less than or equal to trigger threshold corresponding to service type". After the terminal device starts to transmit the data packet of the AI non-video task to the cloud device through the network device, the terminal device can detect the remaining time length of the non-video task, and once the remaining time length of the non-video task is lower than the trigger threshold corresponding to the service type, the DSR can be triggered immediately, and the remaining byte of the task is reported to the network device, so that the network device reserves resources in advance for the task according to the remaining data amount. In this way, when the data of the task arrives in a scattered manner, it can directly fall on the reserved resources, without the need to trigger a scheduling request again, solving the queuing delay problem and improving the scheduling efficiency and transmission reliability.

[0010] In a possible implementation, the first AI non-video task belongs to a first type of AI non-video task, and different types of AI non-video tasks correspond to different trigger time thresholds. In the above scheme, the AI non-video task is an AI data interaction task generated based on discrete events or instructions, and the interaction task relies on non-continuous and discrete data packets, such as voice data, image data, text data, and sensor data. For different types of AI non-video tasks, the data packet size is different, and the AI server also has different inference times according to the data packet. Therefore, the network device can set different trigger time thresholds for various AI non-video tasks according to their characteristics.

[0011] In a possible implementation, before starting to send the data packet of the first AI non-video task to the cloud device through the network device, the method can further include: setting a time length of a PDCP discard timer of the first AI non-video task to infinity. In the above scheme, the PDCP discard timer is set to infinity, which is equivalent to disabling the PDCP discard timer. In this way, the data packet of the task will not be discarded when the PDCP discard timer does not time out, and therefore there is no delay problem caused by retransmitting the data packet.

[0012] In a possible implementation, in a case where the remaining duration of the first AI non-video task is less than or equal to the first trigger time threshold, sending the DSR to the network device can include: obtaining the remaining data amount and the remaining duration of the first AI non-video task whenever the sending window of the first AI non-video task is updated; and in a case where the remaining duration of the first AI non-video task is less than or equal to the first trigger time threshold and the first duration is greater than or equal to the anti-jitter timing value, sending the DSR to the network device, the first duration being a duration from a time when the DSR is last sent to the network device to a time when the sending window is updated. In the foregoing scheme, the terminal device triggers local calculation once immediately whenever the sending window of the terminal device is updated, to obtain the remaining data amount and the remaining duration of the current task. In addition, by setting the anti-jitter timing value, it is possible to avoid resource occupation caused by frequent sending of the DSR to the network device in a short time.

[0013] In a possible implementation, obtaining the remaining data amount can include: determining the remaining data amount according to an estimated data amount of the first AI non-video task and a data amount that has been confirmed by the cloud device. For example, a difference between the estimated data amount totalBytes of the first AI non-video task and the data amount ackedBytes that has been confirmed by the cloud device can reflect how much data of the task is left to be sent.

[0014] In a possible implementation, obtaining the remaining duration of the first AI non-video task can include: determining the remaining duration of the first AI non-video task according to a maximum delay allowed by the first AI non-video task and a duration consumed from a time when a data packet of the first AI non-video task is sent to a time when the sending window is updated. For example, a difference between the maximum delay deadline allowed by the first AI non-video task and the duration T elapsed consumed from the time when the data packet of the first AI non-video task is sent to the time when the sending window is updated can reflect how much time of the task is left to exceed the maximum delay allowed by the task.

[0015] In a possible implementation, the method can further include: receiving a radio resource control reconfiguration message from the network device. The radio resource control reconfiguration message can include the anti-jitter timing value and the trigger time threshold corresponding to the plurality of AI non-video tasks. Different types of AI non-video tasks correspond to different trigger time thresholds. In the above scheme, the RRC reconfiguration message is an instruction sent by the network device to the terminal device, which is used to dynamically modify the configuration parameters of the terminal device on the basis of the established RRC connection, so as to adapt to the changing wireless environment, network state and service demand. By setting the anti-jitter timing value, the terminal device can avoid continuously reporting the DSR in a short time. In addition, by setting the trigger time threshold corresponding to each task in the plurality of tasks, when the remaining time length of a certain type of AI non-video task being executed by the terminal device is less than the trigger time threshold corresponding to the task, the DSR is triggered.

[0016] In a possible implementation, the DSR can further include the remaining time length of the first AI non-video task, and the remaining time length of the first AI non-video task and the remaining data amount are used together to trigger the network device to reserve resources. In the above scheme, by reporting the remaining time length and the remaining data amount to the network device, the terminal device can make the network device more accurately calculate the reserved resources for the remaining data amount.

[0017] In a possible implementation, the method can further include: in a case where the number of resources required by the data packet to be transmitted in the first AI non-video task exceeds the number of resources indicated by the first uplink grant, sending a short buffer status report to the network device, the short buffer status report can include a first index value, and the first index value is used to trigger the network device to additionally reserve resources. Receiving a second uplink grant from the network device. Continue to send the data packet to be transmitted in the first AI non-video task to the cloud device through the network device on the resources additionally reserved by the second uplink grant. In the above scheme, the terminal device can use the resources indicated by the uplink grant to send the remaining data segments. When new data arrives subsequently and the uplink grant is insufficient, the terminal device can initiate a resource addition request to the network device through the short-BSR-remaining field in the MAC PDU, so that the network device can calculate additional resources according to the short-BSR-remaining field information and issue an additional uplink grant. By adding resources for this service, the newly arrived data can fall on the additional resources, without the need to trigger a scheduling request again, thereby solving the queuing delay problem and improving the scheduling efficiency and transmission reliability.

[0018] In a possible implementation, the buffer of the short buffer status report is composed of P bits, Q bits of the P bits are used to write index values, different index values are used to append different numbers of resources, and P and Q are positive integers. For example, P = 5 and Q = 3. In the foregoing scheme, the terminal device and the network device can agree on the number of resources corresponding to each level of appended resource index in advance. For example, as the index number value of the appended resource index increases, the number of resources corresponding to each level of appended resource index increases linearly or exponentially.

[0019] In a possible implementation, the sending of the short buffer status report to the network device can include: sending a MAC PDU to the network device, the MAC PDU can include a first subheader, a second subheader, the short buffer status report, and a MAC SDU next to the short buffer status report, the logical channel identifier of the first subheader points to the short buffer status report, the logical channel identifier of the second subheader points to the MAC SDU, and the MAC SDU can include a next data packet to be sent. In the foregoing scheme, the terminal device can carry the short BSR when sending a data packet, thereby saving signaling overhead.

[0020] In a second aspect, a DSR transmission method is provided. The method can be applied to a network device. The method can be executed by the network device, or can also be executed by a component (such as a circuit, a chip, or a chip system, etc.) configured in the network device, and can also be implemented by a logic module or software that can implement all or part of the functions of the network device.

[0021] The method can include:

[0022] The method can include:

[0023] In the foregoing scheme, the DSR from the terminal device carries a remaining data amount of the first AI non-video task. After receiving the DSR, the network device can calculate the number of resources that need to be reserved in advance for the task according to the remaining data amount, so that the remaining data packets of the terminal device can be directly transmitted using the reserved resources, without triggering a scheduling request again, thereby solving the queuing delay problem and improving the scheduling efficiency and transmission reliability.

[0024] In a possible implementation, reserving resources for the first AI non-video task according to the remaining data amount can include two schemes. In the first scheme, the number of resources reserved for the first AI non-video task is determined according to the remaining data amount and the number of bits contained in one transport block. In the second scheme, a minimum instantaneous rate is determined according to the remaining data amount and a remaining time length of the first AI non-video task; the number of bits contained in one transport block is determined according to the minimum instantaneous rate, different minimum instantaneous rates corresponding to different numbers of bits; and the number of resources reserved for the first AI non-video task is determined according to the remaining data amount and the number of bits contained in one transport block. Compared with the first scheme, the second scheme can dynamically adjust the size of the transport block in combination with the remaining time length and the remaining data amount, and thus more accurately calculate the reserved resources for the remaining data amount.

[0025] In a possible implementation, after sending the first uplink grant to the terminal device, the method can further include: receiving a short buffer status report from the terminal device, the short buffer status report can include a first index value; reserving resources for the first AI non-video task according to the first index value; sending a second uplink grant to the terminal device; and continuing to receive data packets of the first AI non-video task from the terminal device on resources indicated by the second uplink grant, and continuing to forward the data packets of the first AI non-video task to the cloud device. In the foregoing scheme, when the resources reserved by the network device for the terminal device through the first uplink grant are insufficient, the network device can reserve additional resources for the terminal device through the second uplink grant. By reserving additional resources, the terminal device can transmit data packets using the additional resources, thereby solving the queuing delay problem, improving the scheduling efficiency and transmission reliability.

[0026] In a possible implementation, the buffer of the short buffer status report is composed of P bits, Q bits of the P bits are used to write an index value, different index values are used to reserve different numbers of resources, and P and Q are positive integers. P=5 and Q=3. In the foregoing scheme, the terminal device and the network device can agree on the number of resources corresponding to each level of additional resource index in advance, for example, the number of resources corresponding to each level of additional resource index increases linearly or exponentially as the index number value of the additional resource index increases. The network device can analyze the index value of the Q bits to calculate the number of resources that need to be reserved.

[0027] In a possible implementation, the method further includes: sending, to the terminal device, a radio resource control reconfiguration message. The radio resource control reconfiguration message includes the anti-jitter timing value and the trigger time threshold corresponding to the plurality of AI non-video tasks, and different types of AI non-video tasks correspond to different trigger time thresholds. In the foregoing solution, by setting the anti-jitter timing value, the terminal device can be prevented from continuously reporting the DSR in a short time. In addition, by setting the trigger time threshold corresponding to each of the plurality of tasks, when the remaining time length of a certain type of AI non-video task being executed by the terminal device is less than the trigger time threshold corresponding to the task, the DSR is triggered.

[0028] In a third aspect, the present application provides a terminal device, which includes one or more processors and a memory. The memory is coupled to the one or more processors, and the memory is configured to store computer program codes including computer instructions. The one or more processors invoke the computer instructions to enable the terminal device to perform the method provided in the first aspect and any possible implementation thereof.

[0029] In a fourth aspect, the present application provides a network device, which includes one or more processors and a memory. The memory is coupled to the one or more processors, and the memory is configured to store computer program codes including computer instructions. The one or more processors invoke the computer instructions to enable the network device to perform the method provided in the second aspect and any possible implementation thereof.

[0030] In a fifth aspect, the present application provides a communication system, which can include a terminal device, a network device, and a cloud device. The terminal device is configured to perform the prediction information reporting method according to any one of the first aspect, the network device is configured to perform the prediction information reporting method according to any one of the second aspect, and the cloud device is configured to complete reasoning by relying on AI technology.

[0031] In a sixth aspect, the present application provides a computer readable storage medium, which stores a computer program. When the computer program runs on a terminal device, the terminal device performs the method provided in the first aspect and any possible implementation thereof. When the computer program runs on a network device, the network device performs the method provided in the second aspect and any possible implementation thereof.

[0032] In a seventh aspect, the present application provides a chip, which is coupled to a memory. The chip is configured to read and execute a computer program stored in the memory, so as to implement the method provided in any one of the first aspect or the second aspect.

[0033] In an eighth aspect, there is provided a computer program product, which, when run on a computer, causes the computer to perform the method provided in any one of the first aspect or the second aspect.

[0034] It can be understood that the beneficial effects of the third aspect to the eighth aspect described above can be referred to the related description in the first aspect and the second aspect, which will not be repeated here. BRIEF DESCRIPTION OF DRAWINGS

[0035] Figure 1 A reporting flow diagram of DSR for AI video traffic provided by the embodiments of the present application;

[0036] Figure 2 A schematic diagram of a communication system based on AI service provided by the embodiments of the present application;

[0037] Figure 3 A reporting flow diagram of DSR for mobile AI video traffic provided by the embodiments of the present application;

[0038] Figure 4 A transmission diagram of AI non-video traffic provided by the embodiments of the present application;

[0039] Figure 5 A DSR reporting flow diagram of AI non-video traffic provided by the embodiments of the present application;

[0040] Figure 6 A flow diagram of a DSR transmission method for AI non-video traffic provided by the embodiments of the present application;

[0041] Figure 7 A schematic diagram of a MAC PDU provided by the embodiments of the present application;

[0042] Figure 8 A flow diagram of adding resources for service provided by the embodiments of the present application;

[0043] Figure 9 Another schematic diagram of a MAC PDU provided by the embodiments of the present application;

[0044] Figure 10 A schematic block diagram of a communication device provided by the embodiments of the present application. DETAILED DESCRIPTION

[0045] The terms "first" and "second" and the like in the description of the present application and in the claims of the present application are used for the purpose of differentiating different objects, or for the purpose of differentiating different treatments of the same object, rather than for the purpose of describing a specific order of the objects. In addition, the terms "comprise", "have" and any variations thereof in the description of the present application are intended to cover the non-exclusive inclusion. For example, a process, method, system, product or device that comprises a series of steps or units is not limited to the listed steps or units, but can optionally further comprise other steps or units that are not listed, or can optionally further comprise other steps or units that are inherent to the process, method, product or device. In the embodiments of the present application, "a plurality of" includes two or more. In the embodiments of the present application, the words "exemplary" or "for example" or the like are used to mean serving as an example, instance, or illustration. In addition, the network architecture and business scenarios described in the embodiments of the present application are for the purpose of more clearly illustrating the technical solutions of the embodiments of the present application, and do not constitute a limitation on the technical solutions provided by the embodiments of the present application. Those skilled in the art can know that, as the network architecture evolves and new business scenarios appear, the technical solutions provided by the embodiments of the present application are also applicable to similar technical problems.

[0046] The technical solutions provided by the embodiments of the present application can be applied to various communication systems, for example, a global system for mobile communications (GSM) system, a general packet radio service (GPRS), a wireless local area network (WLAN), a long term evolution (LTE) system, an LTE frequency division duplex (FDD) system, an LTE time division duplex (TDD), a sidelink communication system, a universal mobile telecommunication system (UMTS), a worldwide interoperability for microwave access (WiMAX) communication system, a non-terrestrial network (NTN) communication system, a 5th generation (5G) mobile communication system or a new radio (NR) system, and the like. The 5G mobile communication system can include a non-standalone (NSA) and / or a standalone (SA). The technical solutions provided by the embodiments of the present application can also be applied to future communication systems, which are not limited herein.

[0047] Figure 1 A schematic diagram of a communication system 100 is provided for the embodiments of the present application.

[0048] The communication system 100 can include a network device, for example, a network device 110 as shown. Figure 1 The communication system 100 can also include a terminal device, for example, a terminal device 120 as shown. The network device 110 and the terminal device 120 can communicate through a wireless link. Figure 1

[0049] An example of one network device 110 and one terminal device 120 is shown. In a possible implementation, the communication system 100 can also include multiple network devices and / or multiple terminal devices. Figure 1

[0050] ​The network device in the embodiments of the present application can be a device on the network side, such as an access network device, a core network device, etc. The access network device is also sometimes referred to as an access node. The access network device has a wireless transceiving function and is used to communicate with a terminal device. The access network device includes, but is not limited to, a base station in the above-mentioned communication system, an evolved NodeB (eNodeB), a transmit / receive Point (TRP), an NR Node B (gNB) in a 5G mobile communication system, a next generation eNodeB (ng-eNB) in a 5G mobile communication system, an access network device in an open RAN (ORAN) system or a module of the access network device, a satellite in a non terrestrial network (NTN) communication system, a base station in a future mobile communication system, or an access node in a wireless fidelity (Wi-Fi) system, etc. The access network device can also be a module or unit capable of realizing part of the function of a base station. The access network device can be a macro base station, a micro base station or an indoor station, a relay node or a donor node, or a wireless controller in a cloud radio access network (CRAN) scenario, etc. The access network device can also be a server, a wearable device or a vehicle-mounted device, etc. Multiple access network devices in a communication system can be the same type of base station or different types of base stations. The base station can directly communicate with the terminal device or communicate with the terminal device through a relay station. The terminal device can communicate with multiple base stations in different access technologies. The embodiments of the present application do not limit the specific technology and specific device form of the access network device.

[0051] In the embodiments of the present application, the device for realizing the function of the network device can be a network device or a device capable of supporting the network device to realize the function, such as a processor, a circuit, a chip or a chip system, etc. The device can be installed in the network device or used in connection with the network device.

[0052] The terminal device in the embodiments of the present application can be a wireless terminal device capable of receiving network device scheduling and indication information. The wireless terminal device can be a device that provides voice and / or data connectivity to a user, or a handheld device with wireless connection function, or other processing devices connected to a wireless modem. For example, the terminal device can communicate with one or more core networks or the Internet through a radio access network (RAN). The terminal device can also be referred to as a terminal, user equipment (UE), mobile station, mobile terminal, etc. The terminal device can be widely used in various scenarios, such as device-to-device (D2D), vehicle-to-everything (V2X) communication, machine-type communication (MTC), internet of things (IoT), ultra-reliable low-latency communication (URLLC), virtual reality, augmented reality, industrial control, autonomous driving, remote medical treatment, smart power grid, smart furniture, smart office, smart wear, smart transportation, smart city, or satellite communication, etc. The terminal device can be a mobile phone, tablet computer, computer with wireless transceiver function, wearable device, vehicle, aircraft (such as unmanned aerial vehicle, helicopter, airplane), hot air balloon, ship, robot, mechanical arm or smart home device, etc. The embodiments of the present application do not limit the form of the terminal device.

[0053] In the 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 processor, circuit, chip or chip system, etc., which can be installed in the terminal device or used in connection with the terminal device.

[0054] The access network device and / or the terminal device can be fixed or mobile. The access network device and / or the terminal device can be deployed on land, including indoor or outdoor, handheld or vehicle-mounted; can also be deployed on the water surface; can also be deployed on aircraft, balloons and artificial satellites in the air. The embodiments of the present application do not limit the application scenarios of the access network device and the terminal device. The access network device and the terminal device can be deployed in the same scenario or different scenarios, for example, the access network device and the terminal device are deployed on land at the same time; or the access network device is deployed on land and the terminal device is deployed on the water surface, etc., which will not be listed one by one.

[0055] Figure 2 A schematic diagram of a communication system based on AI service provided by the embodiments of the present application.

[0056] The communication system can be applicable to 5G evolution (5G-Advanced, 5.5G) networks and 6G networks, etc. The communication system can include functional entities such as terminal devices, access network devices, core network devices, and cloud devices. The respective functional entities communicate with each other through corresponding interfaces to implement AI services.

[0057] The terminal device is responsible for generating uplink data of AI services, such as voice messages input by a user, environmental data collected through a sensor, pictures or videos collected through a camera, etc.; and receiving and processing downlink data, such as AI inference results.

[0058] The access network device is responsible for functions such as resource scheduling of terminal devices, radio resource management, and radio resource control. Taking the next generation radio access network (NG-RAN) as an example, the NG-RAN can include one or more access network devices, such as gNBs. The gNB is a node that provides NR user plane and control plane protocols to the terminal device, such as a TRP. The access network device can communicate with the core network device through wired or wireless means, such as connecting to the core network through an NG-C interface. As an example, the access network device can allocate uplink grant (UL grant) resources for uplink data from the terminal device, and is responsible for data encryption, segmentation, retransmission, etc.

[0059] The core network device is mainly used for network management, control, and data transmission. The core network device can include user plane functions (UPFs) and control plane functions. The UPF is responsible for data routing and forwarding, executing control plane issued policies, and data packet inspection and charging, etc. The control plane function is responsible for access and mobility management, session management, and formulating policy rules, etc., such as the access and mobility management function (AMF), the session management function (SMF), and the location management function (LMF).

[0060] The cloud device includes various servers located in a data center, such as AI servers. The AI server is responsible for AI inference and training, and can provide powerful AI computing power to process requests sent by the terminal.

[0061] Take the AI image recognition function of a user using a mobile phone as an example. The data flow path is as follows: the user uses the mobile phone to take a photo, and the application packages the picture. The mobile phone sends the data packet to the nearest base station through a wireless signal. The base station sends the data packet to the user plane function (UPF) through the N3 interface. The UPF routes the data packet to the designated AI cloud server through the N6 interface according to the policy. Throughout the process, the control plane functions such as AMF and SMF ensure that the connection is always valid and seamlessly switches when the user moves. The AI server receives the picture, runs the recognition algorithm, and generates the inference result, such as "this is a cat". Then, the inference result returns along the original path and is finally displayed on the screen of the mobile phone.

[0062] To facilitate understanding of the embodiments of the present application, the terms involved in the embodiments of the present application are briefly explained below. The explanation of some terms can refer to the explanation in the 3rd generation partnership project (3rd generation partnership project, 3GPP) standard protocol. It should be understood that the technical terms in the embodiments of the present application are only used as examples and are not limited. For example, as technology evolves, technical terms may also change, and in the case of the same technical meaning, other technical terms should also apply to the present application.

[0063] Extended reality (extended reality, XR) enhancement mechanism refers to a complete set of cross-layer optimization solutions introduced for virtual reality (virtual reality, VR), augmented reality (augmented reality, AR) and mixed reality (mixed reality, MR) technologies in 5.5G networks and 6G networks. Through seamlessly integrating virtual information such as digital images, 3D models, sounds, and tactile feedback with the real world, natural and real-time interaction between users and these virtual-real combined content can be achieved.

[0064] Traditional XR enhancement mechanisms, such as protocol data unit set (protocol data unit set, PDU set), delay status report (delay status report, DSR) based on packet data convergence protocol (packet data convergence protocol, PDCP) discard timer, and priority fallback, are all designed for real-time video traffic.

[0065] In the framework of traditional XR enhancement mechanisms, real-time videos can be divided into two categories: downlink real-time videos and uplink real-time videos. Downlink real-time videos are responsible for delivering virtual content to users, mainly including video pass-through streams, block / tetrahedron-dependent streams, etc. Uplink real-time videos are responsible for sending user's environment information to cloud devices or edge servers, mainly including perception and positioning video streams, encoded video streams, etc.

[0066] Taking the perception and positioning video stream as an example, in combination with Figure 3 , a DSR reporting process for mobile AI video traffic is introduced. Mobile AI refers to AI applications that occur on mobile devices such as mobile phones, AR glasses, and vehicle terminals, or interact with mobile devices. In order to stably place virtual objects in the real world, mobile AI can continuously capture surrounding environment video and depth information through cameras and depth sensors, and upload these information in the form of video streams to cloud devices such as AI servers through network devices. Cloud devices can perform simultaneous localization and mapping (SLAM) according to these information.

[0067] Exemplarily, Figure 3 A DSR reporting process for mobile AI video traffic provided by an embodiment of the present application is shown in the figure.

[0068] The terminal device obtains a series of PDCP data packets, such as video packet V1, video packet V2, and video packet V3.

[0069] The terminal device starts a PDCP discard timer for each PDCP data packet to be sent.

[0070] Video packet V1 and video packet V3 have been sent, while video packet V2 fails to be sent in time due to underlying retransmission failure or scheduling delay, etc., resulting in out-of-order video packets received by the network device.

[0071] After the PDCP discard timer of video packet V2 expires, if the maximum allowed delay set for the video service is exceeded, the terminal device will discard video packet V2 and remove it from the retransmission buffer.

[0072] When preparing to send the next data packet, the terminal device will embed a special control information in the PDCP header of this data packet, which is a delayed suggestion for sending a status report, referred to as DSR control element (CE) or DSR indication information. The DSR indication information includes the PDCP sequence number of the discarded data packet, such as the sequence number of video packet V2.

[0073] After the network device receives the next data packet carrying the DSR indication information, the PDCP layer of the network device parses the DSR indication information. Once it is confirmed that the terminal device has discarded the video packet V2, the network device will immediately stop the reordering timer and generate a PDCP status report. The PDCP status report contains the sequence numbers of the data packets that have been detected and the sequence numbers of the data packets that have not been received. Then, the network device immediately returns the PDCP status report to the terminal device. The PDCP status report can be a separate PDCP control packet or can be carried on the reverse data packet.

[0074] The terminal device receives the PDCP status report fed back by the network device.

[0075] For the video packet, such as the video packet V2, requested for retransmission in the PDCP status report, the terminal device checks its buffer:

[0076] If the video packet is still in the buffer (i.e., not discarded by the PDCP Discard Timer), retransmission is performed. Usually, since the discard timer of the video packet V2 has expired, the video packet V2 is not in the buffer.

[0077] If the video packet, such as the video packet V2, has been discarded, the terminal device ignores the retransmission request for the packet.

[0078] For mobile AI video traffic, the video encoder usually outputs data at a fixed or variable code rate, and the video decoder can usually handle a certain degree of frame reordering, but is extremely sensitive to delay and can tolerate a certain degree of data loss to ensure the smoothness of the picture, so it can choose the user datagram protocol (UDP) or the real-time transport protocol (RTP). Based on this, the terminal device can inform the server that the video frame is incomplete through the application layer. The server has a certain fault tolerance capability. In the case where it is known that the data packet has been lost, the server can still perform degraded inference based on the received incomplete frame data.

[0079] With the rise of AI applications, various LLMs, AI-driven image processing tools, and autonomous driving applications, the user experience of mobile devices such as smartphones, smart glasses, smart watches, and connected vehicles has been greatly enhanced. These mobile AI applications can not only provide AI services for videos, but also for non-video such as voice question and answer, image upload, and text interaction. For AI data interaction scenarios other than continuous, high-code-rate video streams (hereinafter referred to as AI non-video transmission), in order to enable the AI on the cloud device to react in a similar way to humans, the total delay of the AI process needs to be limited to a very short time, for example, 300 milliseconds. However, AI inference usually consumes most of the time, such as more than 200 milliseconds, leaving only a small delay budget for network communication between the terminal device and the cloud device. In order to ensure the deterministic user experience of real-time interactive mobile AI services originating from mobile devices, it is crucial to strengthen the delay guarantee.

[0080] Mobile AI video is transmitted based on a protocol without congestion control, allowing some data loss, and the AI server can perform degraded inference based on the received, incomplete frame data. Unlike mobile AI video, mobile AI non-video traffic transmission is based on reliable transmission protocols with congestion control, such as Transmission Control Protocol (TCP) / Quick UDP Internet Connections (QUIC), to ensure that every byte arrives at the receiving end completely and in order, so that the server can perform AI inference based on complete data.

[0081] For ease of understanding, the following takes voice question and answer, image upload, and text interaction as examples to explain why mobile AI non-video traffic chooses reliable transmission protocols with congestion control.

[0082] As an example of a scenario, a user speaks a sentence to a terminal device, and a smart assistant uploads the sentence to a cloud device, and the cloud device returns a segment of voice or text corresponding to the sentence. For the voice question and answer scenario, this kind of data is bursty, not continuous stream, and the loss of a packet can cause voice interruption or recognition error.

[0083] As another example of a scenario, a user can use a mobile phone to take pictures, recognize objects, pay by scanning codes, or upload pictures for AI photo editing, and the AI application uploads these pictures to a cloud device, and the cloud device returns the recognition result, payment result, or processing result corresponding to the image. For the image upload scenario, this kind of data is a single or small number of large files, and data loss during image upload will cause the picture to be damaged and the AI to be unable to analyze correctly.

[0084] As another example of a scenario, a user can use a terminal device to have a dialogue with a large language model, intelligent customer service, and the terminal device uploads the sentence to a cloud device. For a text interaction scenario, the data is a short message. A lost data packet can cause the AI model to understand incorrectly or generate an incorrect output.

[0085] A reliable transmission protocol based on congestion control refers to a class of network protocols that simultaneously achieve the following two goals through a set of complex mechanisms: 1. Reliability, ensuring that each byte sent from the sending end is complete, in order, and reaches the receiving end without loss, duplication, or error. AI processing usually relies on the complete context of the data. If the lost data or the out-of-order data arrives, it can cause the AI to be unable to be correctly parsed. 2. Congestion control. The mobile network environment is complex and variable, such as cellular signal fluctuations or wireless fidelity (Wi-Fi) switching. If there is no congestion control, the AI application can quickly retransmit when the network is poor, exacerbating congestion. The sending end can avoid network overload crashes by dynamically sensing the congestion in the network and actively adjusting the sending rate of the data packet, and can achieve fair sharing of bandwidth between conflicting data streams.

[0086] Exemplary, Figure 4 A transmission schematic of AI non-video traffic provided by an embodiment of the present application is shown.

[0087] As Figure 4 shown, the terminal device can use a layered architecture to divide the software into several layers, each layer having a clear role and division of labor. Layers communicate with each other through software interfaces. In some embodiments, the software layers of the software architecture are divided from top to bottom as follows: application layer, transport layer, internet protocol (IP) layer, and access layer.

[0088] The application layer is responsible for generating data and initiating system calls. The application layer is the highest layer, and voice, image, text, and other AI non-video data are generated at this layer. The application layer organizes the data, user identification (ID), request type (such as “image recognition”), and other information according to the agreed format (such as JSON, protocol buffers) with the AI server to form a complete application layer message. Then, the application program initiates a network request through the socket application programming interface (API) provided by the operating system, which specifies the target server IP address and port number and transmits the prepared application layer message.

[0089] The transport layer is responsible for end-to-end reliable data transmission, e.g., using TCP or QUIC, etc. protocols with congestion control. Take TCP as an example, after receiving the application layer message from the socket at the TCP layer, TCP will perform the following operations: if the application data is too large, it will be segmented into blocks of the maximum segment size (MSS) suitable for network transmission; add a TCP header to each data segment, such as source and destination port numbers, sequence numbers, acknowledgment numbers, window sizes, and a series of flag bits, etc. At this time, the data unit is called a TCP segment. In addition, the congestion control algorithm of TCP starts to work, which dynamically adjusts the congestion window according to network feedback to determine how many unacknowledged data segments can be sent to prevent network overload and ensure reliable transmission.

[0090] The IP layer is responsible for addressing and routing of data packets. At this layer, the overall data of the mobile AI non-video is divided into multiple smaller packets. For example, after receiving the TCP segment issued by the TCP layer at the IP layer, the IP layer will add an IP header to each TCP segment to form an IP packet, and the IP header can include the source IP address, the destination IP address, the protocol number, and the time to live. The operating system can determine the next hop of the IP packet according to the destination IP address and the local routing table. For packets sent to the Internet, the next hop is usually the default gateway.

[0091] The access layer is the lower layer of wireless link processing part, where the data packet (i.e., IP packet) is queued and buffered, waiting to be sent through the air interface. This layer manages the physical transmission of data packets to ensure that data can enter the wireless channel. The access layer can include, from top to bottom, a service data adaptation protocol (SDAP) layer, a PDCP layer, a radio link control (RLC) layer, a media access control (MAC) layer, and a physical layer. Among them, the SDAP layer is responsible for the mapping between QoS (quality of service) flow and data radio bearer. The PDCP layer is responsible for security, integrity protection, and duplicate data deletion, e.g., enabling DSR based on PDCP discard timer, starting the timer for the data packet. The RLC layer is responsible for segmentation / concatenation and automatic retransmission. The MAC layer is responsible for scheduling, multiplexing, and hybrid automatic repeat request. The physical layer is responsible for the final processing of digital signals and the transmission of wireless signals.

[0092] In the congestion control mechanism of the above-mentioned transport layer, the sending end maintains a congestion window. The congestion window represents how much data the sending end can inject into the network without receiving an acknowledgment (ACK) message from the receiving end. The congestion algorithm can include the following stages: at the beginning of the connection, the congestion window starts from a small value (such as 1 MSS), exponentially increases, and quickly detects the available bandwidth; when the window reaches a threshold, it changes to linear growth; when a packet loss (timeout or repeated ACK) is detected, it is determined that the network has been congested, and the window can be quickly reduced (such as halved or reset to 1) to reduce network load; after fast retransmission, instead of entering slow start directly, a certain rate is maintained to smooth performance.

[0093] Based on the description of the above embodiments, the DSR based on the PDCP discard timer is designed for video traffic, and AI video is transmitted based on a protocol without congestion control (such as UDP), so in the DSR mechanism, when the PDCP discard timer of the sending end times out, the sending end does not necessarily retransmit the data packet after receiving the PDCP status report by sending the DSR. However, AI non-video traffic is transmitted based on a reliable transmission protocol with congestion control (such as TCP or QUIC), and if the DSR mechanism is directly applied to AI non-video traffic, the data packet will be discarded when the PDCP discard timer of the sending end times out, triggering the DSR mechanism. In order to ensure the reliability of the data, the sending end must retransmit the data packet, which will inevitably increase the delay.

[0094] An exemplary AI non-video traffic DSR reporting process is provided. Figure 5 An exemplary AI non-video traffic DSR reporting process is provided.

[0095] After the application layer of the terminal device obtains AI non-video data (such as a picture), the application layer generates a complete application layer message according to the format agreed with the server, and initiates a network request through an API.

[0096] The transport layer of the terminal device divides the application layer message into multiple data packets / blocks of MSS size suitable for network transmission, and adds a TCP header to each data segment. The congestion control algorithm of the transport layer starts to work, and it dynamically adjusts the congestion window according to network feedback to determine how many data segments can be sent to prevent network overload.

[0097] The data packet reaches the access layer of the terminal device after being processed by the IP layer. The access layer of the terminal device starts a PDCP discard timer for each data packet. For example, after data packet N reaches the PDCP layer from the upper layer, the PDCP layer assigns a sequence number to data packet N, and performs encryption and integrity protection to form a PDCP PDU. The PDCP layer starts a PDCP discard timer for data packet N, and the duration of the PDCP discard timer is usually configured by radio resource control (RRC) signaling according to the QoS characteristics of the bearer. The PDCP PDU is delivered to the RLC layer for subsequent processing. At this time, the state of the PDCP PDU in the terminal device is “sent but not confirmed”. Before the timer expires, if an acknowledgment message from the PDCP layer of the receiving end is received, the PDCP layer of the terminal device will stop the discard timer of data packet N and clear its context. If the discard timer expires, regardless of whether the RLC layer of the lower layer is attempting retransmission, the PDCP layer will immediately perform a discard operation and clear it from the retransmission buffer.

[0098] The terminal device sends DSR indication information to the network device. The DSR indication information includes the PDCP sequence number of the discarded data packet, such as the sequence number of data packet N.

[0099] After the network device receives the DSR indication information, the PDCP layer of the network device parses the DSR indication information. Once it is confirmed that data packet N has been discarded by the terminal device, the network device generates a PDCP status report. The PDCP status report includes the sequence numbers of the data packets that have been detected and the sequence numbers of the data packets that have not been received. Then, the network device returns the PDCP status report to the terminal device.

[0100] For AI non-video services, a missing data packet can cause the AI model to understand incorrectly or generate incorrect output, so it is necessary to ensure that each byte reaches the receiving end completely and in order. After determining that data packet N is missing according to the PDCP status report, the terminal device determines to retransmit data packet N.

[0101] In the above process, from discarding data packet N by the terminal device to retransmitting data packet N, a certain amount of time is required, which causes other data packets after data packet N to be unable to be successfully sent out, resulting in congestion. In addition, the loss of data packet N not only causes the PDCP discard timer of data packet N to expire, but also triggers the congestion control algorithm of the transport layer to misjudge that the network is congested, thereby greatly reducing the congestion window, so that subsequent data packets are all blocked by the congestion window, causing the delay to increase sharply.

[0102] In summary, for AI non-video traffic based on a reliable transmission protocol with congestion control, any form of packet discard will increase the latency, so the traditional DSR mechanism cannot be directly used for AI non-video traffic and needs to be adapted and enhanced. To avoid packet discard, the PDCP discard timer can be disabled, but after disabling the PDCP discard timer, the traditional DSR trigger condition is completely invalid.

[0103] In addition, for AI non-video traffic, when the data packets arrive at the network device in a scattered manner, the network device can only allocate resource blocks (RBs) according to the size of the current data packet, and the subsequent arriving data packets need to be queued again, resulting in RB fragmentation and increased queuing delay, causing inefficient scheduling or unreliable transmission.

[0104] In view of the above problems, the present application provides a DSR transmission method for AI non-video traffic. The method defines a new DSR trigger condition. When the terminal device updates the sending window, the terminal device calculates the remaining bytes and the remaining duration of the current task in real time. Once the remaining duration is lower than the trigger threshold corresponding to the service type, the terminal device immediately triggers DSR and reports the remaining bytes and the remaining duration of the task to the network device. The network device reserves resources in advance for the remaining bytes of the task according to this information.

[0105] In the above method, different types of non-video traffic services correspond to different trigger thresholds. When the remaining duration of a certain type of non-video traffic service is short, triggering DSR can make the network device reserve resources in advance for the remaining bytes of the service. Since the terminal device does not discard data packets, there is no delay problem caused by retransmitting data packets. In addition, when the data of the service arrives in a scattered manner, it can directly fall on the reserved resources, without the need to trigger a scheduling request again, solving the queuing delay problem and improving the scheduling efficiency and transmission reliability.

[0106] It should be noted that when the DSR transmission method provided by the embodiments of the present application is applied to a mobile terminal, the above AI non-video traffic is also referred to as mobile AI non-video traffic.

[0107] The scheme provided by the embodiments of the present application will be described in detail below in conjunction with the corresponding flowcharts. It can be understood that the main body of the interaction in the illustrative flowcharts provided by the embodiments of the present application is taken as an example to illustrate the method, but the present application does not limit the main body of the interaction. For example, the device in the illustrative flowchart can also be a chip, a chip system, or a processor supporting the device to implement the method, and can also be a logic module or software capable of implementing all or part of the functions of the device.

[0108] As a general statement, the message or signaling interactions involved in the interaction process of this application embodiment can be standard messages or signaling or newly introduced messages or signaling. This application embodiment does not make specific limitations on this.

[0109] Figure 6 This is a flowchart illustrating a DSR transmission method for non-video AI traffic according to an embodiment of this application. It can be understood that... Figure 6 The term "terminal device" can refer to a terminal device or a device within a terminal device (such as a processor, chip, or chip system). Figure 6 The network devices mentioned can be network devices, such as access network devices or core network devices, or they can refer to devices within network devices (such as processors, chips, or chip systems). Figure 6 The cloud devices mentioned can refer to cloud devices, such as AI servers, or devices within cloud devices (such as processors, chips, or chip systems).

[0110] like Figure 6 As shown, the method may include the following S01 to S07.

[0111] S01, the network device sends an RRC reconfiguration message to the terminal device.

[0112] An RRC reconfiguration message is a command sent by a network device to a terminal device in the Radio Resource Control (RRC) protocol. It aims to dynamically modify the terminal device's configuration parameters based on an established RRC connection to adapt to constantly changing wireless environments, network conditions, and service requirements. An RRC reconfiguration message is a structured container containing several optional configuration items, such as measurement configuration information, mobility control information, NAS message container, radio bearer configuration, secondary cell configuration, and security configuration.

[0113] In some embodiments, the RRC reconfiguration message may include:

[0114] The DSR trigger time threshold list is dsr_ThresholdList. This list can include trigger time thresholds corresponding to each of multiple tasks, with different trigger time thresholds for different types of AI non-video tasks. It's understood that for different types of AI non-video tasks, the packet size varies, and the time the AI ​​server takes to infer based on the packets also varies. Therefore, network devices can set different trigger time thresholds for various AI non-video tasks based on their characteristics. When the remaining duration of the current task being executed by the terminal device is less than the trigger time threshold corresponding to that task in this list, the terminal device can trigger a DSR.

[0115] Exemplarily, in the trigger time threshold list of the DSR, an index value of a task can be represented by N bits, and different index values correspond to different trigger time thresholds. N is a positive integer, for example, N = 3.

[0116] Table 1 below shows trigger time thresholds corresponding to picture inference, voice assistant, text interaction and health analysis respectively.

[0117] Table 1

[0118]

[0119] The AI non-video tasks and corresponding trigger time thresholds provided in Table 1 above are only exemplary. In actual implementation, the trigger time threshold list of the DSR can include more types of AI non-video tasks, and the trigger time threshold corresponding to each task can be adjusted according to actual needs.

[0120] In some other embodiments, the RRC reconfiguration message can further include:

[0121] A jitter prevention timer dsr-ProhibitTimer. When the remaining duration of the current task being executed by the terminal device is less than the trigger time threshold corresponding to the task, the terminal device can trigger the DSR. When the terminal device sends the DSR to the network device, it occupies a certain air interface resource. If the terminal device frequently sends the DSR to the network device in a short period of time, it may squeeze the resources for processing normal user services and data scheduling requests, and in severe cases, it may cause the base station control plane to overload, the response to slow down or even service interruption. By setting the jitter prevention timer, the terminal device can be prevented from continuously reporting the DSR in a short period of time.

[0122] S02, the terminal device acquires an AI non-video task (referred to as a first AI non-video task).

[0123] The AI non-video task described above is an AI data interaction task generated based on discrete events or instructions, and the interaction task relies on non-continuous, discrete data packets, such as voice data, image data, text data, and sensor data. Among them, the voice data can include voice instructions, voice questions and answers, or real-time translation, etc.; the image data includes uploading of single or small number of pictures for identification, analysis, retouching, etc.; the text data includes natural language dialogue, instruction issuing, code generation, or intelligent customer service, etc.; the sensor data includes structured or semi-structured data from IoT devices, such as temperature, coordinates, device status. Generally, the data transmission of the AI non-video task is in a "request-response" mode, rather than a continuous stream, and there is a clear silent period between two interactions. The AI non-video task is based on a reliable transmission protocol with congestion control, such as TCP / QUIC, to ensure that each byte can be completely and orderly delivered to the receiving end, so that the server can perform AI inference based on complete data.

[0124] In some embodiments, the AI non-video task described above can be triggered by a user. For example, the user is using a real-time AI image processing application in the mobile phone, such as an application for real-time object recognition. The user clicks the "start analysis" control, and the mobile phone obtains a picture inference task to upload the picture data captured by the camera to the AI server for real-time AI inference, and then returns and displays the object recognition result on the screen of the mobile phone.

[0125] In other embodiments, the AI non-video task described above can be triggered by an application. For example, a health application in the mobile phone can periodically report user's heart rate, step count, and exercise path data to the cloud, and the AI server in the cloud analyzes the exercise effect and health risk and generates a report to return to the mobile phone. The health application generates a health data analysis task each time the data is reported.

[0126] S03, the terminal device sets the duration of the PDCP discard timer of the task to infinity (∞).

[0127] Each time the user or the application triggers an AI non-video task, the terminal device sets the PDCP discard timer to ∞, thereby disabling the PDCP discard timer. When the PDCP discard timer is disabled, the PDCP discard timer does not timeout, so that the data packets of the task are not discarded or retransmitted.

[0128] S04, the terminal device creates an AI_UplinkCtx context record for the task.

[0129] Each time the user or the application triggers an AI non-video task, the terminal device also creates an AI_UplinkCtx context record for the task. The record can be used to determine whether to trigger DSR.

[0130] In some embodiments, the context record AI_UplinkCtx described above can include:

[0131] 1. A task identification taskID of the task. All data packets belonging to the same task carry the same task identification. The task identification consists of M bits, the first N bits of which are used to represent the task type of the current task. Wherein, M is an integer greater than N. For example, the task identification of the first task is 00100, and the task identification of the second task is 00101. The first 3 bits of the two task identifications represent that the current task is a voice assistant task. The last 2 bits of the task identifications of the two tasks are 00 and 01 respectively, which are used to distinguish the two voice assistant tasks.

[0132] 2. An estimated byte number totalBytes of the task. The estimated byte number of the task can be obtained by the application layer when creating the task. For example, for a picture recognition task, the resolution of the picture is known, and when the AI image processing application creates the recognition task of the picture, the number of bytes to be uploaded can be estimated. For example, for a voice assistant task, the size of the voice message entered by the user into the terminal device is known, and when the voice assistant application creates the question and answer task of the voice message, the number of bytes to be uploaded can be estimated.

[0133] 3. A maximum delay deadline allowed by the task. Different types of AI non-video tasks have different delay tolerances, and AI task complexity and computing resource requirements also differ. Based on these reasons, the terminal device can set different maximum delays for different tasks. As an optional implementation, the maximum delay allowed by a task can be configured by the network device, such as configured by the RRC reconfiguration message of S01 in the above embodiment. As another optional implementation, the maximum delay allowed by a task can be configured by the terminal device. As still another optional implementation, the maximum delay allowed by a task can be specified by a protocol.

[0134] S05, the terminal device starts sending data packets of the task to the cloud device through the network device.

[0135] In the embodiments of the present application, all data packets belonging to the same task carry the same task identification, and the task identification consists of M bits, the first N bits of which are used to represent the task type of the current task. The network device can distinguish between tasks according to the task identification.

[0136] S06, whenever the sending window is updated, the terminal device calculates the remaining data amount (referred to as remaining data amount) and the remaining time length of the current task. The remaining data amount is the data amount of the data packets estimated to be transmitted in the task, such as the remaining bytes.

[0137] The sending window refers to the maximum amount of data that the sending end (e.g., a terminal device) is allowed to send before receiving an acknowledgement (ACK) message from the receiving end (e.g., a cloud device). The actual sending window is equal to the minimum value of the receiving window and the congestion window. The receiving window is the size of the remaining buffer space announced by the receiving end through a window update message (e.g., an ACK message or a STREAM frame acknowledgement message). The congestion window is the amount of data estimated by the sending end according to the network congestion condition, which will not cause network congestion. The sending window is updated, including the update of the receiving window and the update of the congestion window.

[0138] As an example, the process of updating the receiving window is as follows: the application program of the cloud device takes data from the TCP buffer, and the available buffer space of the cloud device becomes larger. The cloud device carries a larger receiving window value when sending the next ACK message. After the terminal device receives this ACK message, the terminal device can immediately update its own sending window, so that more data can be sent.

[0139] As another example, the process of updating the congestion window is as follows: every time the terminal device receives an ACK message, the congestion control algorithm can confirm that the network condition is good and increase the sending rate tentatively. In this way, in the slow start phase, the congestion window increases exponentially in each round trip time (RTT), such as increasing an MSS every time an ACK message is received. When the congestion window exceeds the slow start threshold, it enters the linear growth phase, and the congestion window increases by an MSS in each RTT.

[0140] Every time the TCP sending window or the QUIC sending window of the terminal device is updated, the terminal device will immediately trigger a local calculation to obtain the remaining bytes and the remaining duration of the current task.

[0141] As an example, the above-mentioned remaining bytes can be calculated by the following relationship:

[0142] Remain_Bytes = totalBytes - ackedBytes.

[0143] Wherein, Remain_Bytes represents the remaining bytes; totalBytes represents the estimated byte number of the task, which can be obtained through the context record AI_UplinkCtx of the above-mentioned embodiment S04; ackedBytes represents the number of bytes that have been sent and confirmed by the receiving end. It can be understood that the difference between totalBytes and ackedBytes can reflect how much data is left to be sent for the task.

[0144] For example, the remaining time length can be calculated by the following relationship:

[0145] Remain_Time = deadline - T_elapsed.

[0146] Wherein, Remain_Time represents the remaining time length; deadline represents the maximum delay allowed for the task, which can be obtained by the context record AI_UplinkCtx of the above-mentioned embodiment S04; T_elapsed is a local timer newly added by the terminal device, which is used to record the time consumed from the start of the task to the current time. It can be understood that the difference between deadline and T_elapsed can reflect how much time the task has left to exceed the maximum delay allowed for the task.

[0147] S07, the terminal device judges whether the remaining time length of the current task is less than or equal to the trigger time threshold corresponding to the current task.

[0148] According to the description of the above-mentioned embodiment S01, the network device can send a list of trigger time thresholds of DSR dsr_ThresholdList to the terminal device through the RRC reconfiguration message, which includes the trigger time threshold corresponding to each task in multiple tasks. According to the description of the above-mentioned embodiment S04, when the terminal device obtains a task, it can create a context record for the task, which includes the task identification taskID of the task, and the first N bits of the task identification are used to represent the task type of the current task. Based on this, the terminal device can retrieve the trigger time threshold corresponding to the current task in dsr_ThresholdList according to the first N bits of the taskID of the current task, and judge whether the remaining time length of the current task is less than or equal to the trigger time threshold corresponding to the current task.

[0149] If the remaining time length of the current task is greater than the trigger time threshold corresponding to the current task, it means that the remaining time length left for the task is sufficient, and then the terminal device can return to execute the above-mentioned S06.

[0150] If the remaining time length of the current task is less than or equal to the trigger time threshold corresponding to the current task, it means that the remaining time length left for the task is insufficient. In an optional implementation, the terminal device can execute the following S10. In another optional implementation, the terminal device can skip the following S08 and S09, and directly execute S10.

[0151] S08, the terminal device obtains the time length from the time of the last trigger DSR to the current time, which is called the first time length.

[0152] S09, the terminal device determines whether the first time length is greater than or equal to the anti-shake timing value dsr-ProhibitTimer.

[0153] If the first time length is less than the anti-shake timing value dsr-ProhibitTimer, it means that the distance from the last triggering of DSR is short. In combination with the description of the above embodiment S01, the terminal device frequently sends DSR to the network device in a short time, which may squeeze the resources for processing normal user services and data scheduling requests, and in severe cases, it may cause the base station control plane to be overloaded, the response to be slow, and even service interruption, so it is necessary to avoid continuous reporting of DSR in a short time. The terminal device can return to execute the above S06.

[0154] If the first time length is greater than or equal to the anti-shake timing value dsr-ProhibitTimer, it means that the distance from the last triggering of DSR is long, and then the terminal device can execute the following S10.

[0155] S10, the terminal device sends DSR to the network device.

[0156] Correspondingly, the network device receives the DSR from the terminal device.

[0157] In some embodiments, the terminal device can send the DSR to the network device through a MAC control element (CE). In an optional implementation, the DSR carries the remaining bytes of the task. In another optional implementation, the DSR carries the remaining bytes and the remaining time length of the task.

[0158] Exemplarily, Figure 7 A schematic diagram of a MAC PDU provided by an embodiment of the present application. The MAC PDU is a basic data block processed and delivered by the MAC layer between logical channels and transport channels. As shown in the figure, Figure 7 The MAC PDU is composed of a MAC header, zero or at least one MAC service data unit (SDU), zero or at least one MAC CE, and optional padding (Padding). The MAC header is composed of multiple sub-headers, each of which corresponds to a MAC SDU or a MAC CE and contains control information, such as a logical channel identifier (LCID) used to indicate which logical channel or which type of MAC CE the data block belongs to. In S10, the terminal device can use one or more MAC CEs to indicate the remaining bytes and the remaining time length of the task; and use a MAC SDU to write user data of the upper layer (such as the RLC layer), for example, a data packet to be sent. That is, the terminal device can carry the DSR when sending a data packet.

[0159] S11, the network device parses the DSR and calculates the number of resources required for transmitting the remaining bytes.

[0160] In an alternative implementation, if the DSR carries the remaining bytes of the task but not the remaining time length, the network device can calculate the number of resources required for transmitting the remaining bytes according to the remaining bytes of the task.

[0161] For example, the network device can calculate the number of resources required for transmitting the remaining bytes according to the following relation 1:

[0162] grant = ceil (Remain_Bytes / TBS).

[0163] wherein grant represents the number of resources required for transmitting the remaining bytes, ceil is a rounding-up operation, Remain_Bytes represents the remaining bytes, and TBS represents the size of a transport block (e.g. the number of bits contained).

[0164] In another alternative implementation, if the DSR carries the remaining bytes and the remaining time length of the task, the network device can calculate the number of resources required for transmitting the remaining bytes according to the remaining bytes and the remaining time length of the task.

[0165] For example, the network device can calculate the minimum instantaneous rate according to the following relation 2:

[0166] Required_Throughput = Remain_Bytes / Remain_Time.

[0167] wherein Required_Throughput represents the minimum instantaneous rate, Remain_Bytes represents the remaining bytes, and Remain_Time represents the remaining time length.

[0168] The scheduler of the network device can dynamically adjust the allocation of resources, i.e. the size of a transport block (TBS), according to the minimum instantaneous rate calculated by the above relation, i.e. if the minimum instantaneous rate is low, the network device can schedule a smaller TBS, and if the minimum instantaneous rate is high, the network device can schedule a larger TBS. Then, the network device can calculate the number of resources required for transmitting the remaining bytes according to the above relation 1.

[0169] S12, the network device issues an uplink grant (UL Grant) (referred to as a first uplink grant) to the terminal device.

[0170] In some embodiments, the network device can send an uplink grant encapsulated by a downlink control information (DCI) format to the terminal device through a physical downlink control channel (PDCCH) channel.

[0171] For example, the first uplink grant can include at least one of the following:

[0172] The uplink grant flag UL-grant-flag indicates that this message is used to inform the resources reserved for the task.

[0173] The hybrid automatic repeat request (HARQ) process ID (Process ID, PID) indicates which HARQ process the current transmission belongs to.

[0174] The time domain resource allocation, such as the slot offset, the starting symbol and the length.

[0175] The frequency domain resource allocation, such as the starting position and the length of the RB. The length of the RB is obtained by the above embodiment S11.

[0176] The modulation and coding scheme (MCS) is a predefined table index corresponding to a specific modulation method (such as QPSK, 16QAM, 64QAM or 256QAM) and coding rate.

[0177] The transmit power control (TPC) is used to adjust the uplink transmit power of the terminal device. For example, the network device can dynamically instruct the terminal device to increase, decrease or maintain the current transmit power according to the measured uplink signal quality (such as path loss and interference), to ensure that the signal reaches the network device with the lowest necessary power and reduce interference.

[0178] S13, the terminal device sends the data packet corresponding to the remaining bytes to the network device on the resource indicated by the uplink grant.

[0179] In the method provided in the application, the duration of the PDCP discard timer is set to infinity, so that the TCP / QUIC layer of the cloud device will not trigger retransmission due to air interface packet loss, thereby increasing the delay. The DSR trigger condition is changed from "PDCP discard timer timeout" to "task remaining duration less than or equal to the trigger threshold corresponding to the service type", so that whenever the terminal device updates the sending window, if the remaining duration is lower than the trigger threshold corresponding to the service type, the terminal device immediately triggers the DSR and reports the remaining bytes and the remaining duration of the task to the network device, so that the network device can reserve resources in advance for the remaining bytes of the task according to the information. When the data arrives, it can directly fall on the reserved resources, without the need to trigger the scheduling request again, thereby solving the queuing delay problem and improving the scheduling efficiency and transmission reliability.

[0180] The above embodiments introduce an implementation method of reserving resources in advance for the service by extending the DSR MAC CE to report the remaining bytes and the remaining duration. According to the description of the above embodiment S11, the number of reserved resources is calculated according to the remaining bytes and the remaining duration, and the remaining bytes are calculated according to the estimated byte number of the task. In some scenarios, the actual data flow of the service may exceed the estimated byte number, resulting in that the reserved resources cannot meet the transmission requirements of the service, and additional resources need to be added for the service. In other scenarios, the wireless environment is real-time changing, and the good channel condition set for the reserved resources may not be maintained, resulting in a decrease in signal strength. In this case, in order to guarantee the same data rate, the same data may need to occupy time domain resources, and therefore additional resources need to be added for the service.

[0181] The implementation method of adding additional resources for the service is exemplarily described below.

[0182] Exemplarily, Figure 8 The flowchart of adding additional resources for the service provided in the embodiments of the application is shown.

[0183] As Figure 8 shown, the terminal device can send the DSR to the network device through S10, and the network device can send the uplink grant to the terminal device for the first time through S12, so that the terminal device can send the remaining bytes of the service on the resources indicated by the uplink grant. If the resources indicated by the uplink grant cannot meet the transmission requirements of the service, additional resources can be added for the service through S14 to S19. For the specific implementation method of S01 to S13, refer to the description of the above embodiments, which will not be described here.

[0184] S14, the terminal device determines that the resources indicated by the uplink grant cannot meet the transmission requirements of the service.

[0185] The phrase "the resources indicated by the uplink authorization cannot meet the transmission requirements of this service" means that the resources required for the remaining amount of data to be sent in this service exceed the resources indicated by the uplink authorization.

[0186] S15, the terminal device sends a short buffer status report (short BSR) via MAC CE.

[0187] Accordingly, the network device receives a short buffer status report from the terminal device.

[0188] A Short BSR is a special type of MAC CE used to quickly report the amount of buffer data in a logical channel group to network devices. In this embodiment, the terminal device can request additional uplink authorization in the short-BSR-remaining field of the MAC PDU. The traditional 5-bit buffer size of the short buffer status report is reused, with the first 3 bits used as the short-BSR-remaining field to write the index of the additional resources.

[0189] For example, Figure 9 This is a schematic diagram of another MAC PDU provided in an embodiment of this application. Figure 9 As shown, the MAC header of a MAC PDU includes at least subheader 1 and subheader 2. Subheader 1's LCID indicates that the next payload is a short BSR, and subheader 2 indicates that the next payload is a MAC PDU. The MAC PDU payload consists of a short BSR and a MAC SDU immediately adjacent to it. The MAC SDU is used to write user data to the upper layer (such as the RLC layer), for example, a data packet to be sent. That is, the terminal device can carry the short BSR when sending the next data packet.

[0190] The short BSR consists of 5 bits. The first 3 bits are used as the short-BSR-remaining field, which is used to write the index of the appended resource. For example, bits 4 to 2 correspond to the following values: 000 for level 1 appended resource index, 001 for level 2, 010 for level 3, 011 for level 4, 100 for level 5, 101 for level 6, 110 for level 7, and 111 for level 8. Terminal devices and network devices can pre-determine the number of resources corresponding to each level of appended resource index. For example, as the index number of the appended resource index increases, the number of resources corresponding to each level of appended resource index grows linearly or exponentially.

[0191] It should be noted that the above embodiment is exemplarily described by taking the short-BSR-remaining field occupying 3 bits as an example, and the application is not limited thereto. In actual implementation, the short-BSR-remaining field can occupy more or less bits. It can be understood that the more bits occupied by the short-BSR-remaining field, the finer the additional resource index division granularity, and the more accurate the number of additional resources, which can avoid wasting uplink resources to a certain extent.

[0192] S16, the network device parses the BSR, and determines the number of additional resources required for transmitting the remaining bytes according to the additional resource index.

[0193] Exemplarily, Table 2 below shows the correspondence between the additional resource index and the number of additional resources. n is a positive integer.

[0194] Table 2

[0195]

[0196] S17, the network device sends the additional uplink grant (referred to as the second uplink grant) to the terminal device.

[0197] In some embodiments, the network device can send the additional uplink grant encapsulated in the DCI format to the terminal device through the PDCCH channel.

[0198] Exemplarily, the second uplink grant can include at least one of the following:

[0199] The additional uplink grant identifier Add-UL-grant-flag indicates that this message is an additional resource on the reserved resource, and there is no need to reinitialize the HARQ.

[0200] The HARQ PID indicates which HARQ process the current transmission belongs to. It should be noted that the HARQ PID of the first uplink grant and the second uplink grant is the same, which ensures that the same HARQ process number is used, and the terminal device will not start a new process, saving buffer time.

[0201] The time domain resource allocation recalculated according to the short-BSR-remaining field.

[0202] The frequency domain resource allocation recalculated according to the short-BSR-remaining field.

[0203] The MCS recalculated according to the short-BSR-remaining field.

[0204] The TPC recalculated according to the short-BSR-remaining field.

[0205] S18, the terminal device sends the data packet corresponding to the remaining bytes to the network device on the resource indicated by the additional uplink grant.

[0206] S19, the network device forwards the data packet from the terminal device to the cloud device in sequence according to the receiving order. The cloud device performs inference based on the received data packet.

[0207] In the above method provided by the present application, the terminal device can send the remaining data segments using the resource indicated by the uplink grant. When there is still new data arriving and the uplink grant is insufficient to cover it, the terminal device can initiate a resource addition request to the network device through the short-BSR-remaining field in the MAC PDU, so that the network device can calculate additional resources according to the short-BSR-remaining field information and issue additional uplink grants. By adding resources for this service, the newly arrived data can fall on the additional resources, without the need to trigger a scheduling request again, solving the queuing delay problem and improving the scheduling efficiency and transmission reliability.

[0208] In order to facilitate understanding of the DSR transmission method provided by the embodiments of the present application, an example of voice interaction is provided below.

[0209] Suppose in an XR service, the terminal device needs to perform voice interaction, and the settings are as follows:

[0210] Task type: voice assistant, and the first three digits of taskID are 001;

[0211] Estimated total number of bytes uploaded totalBytes = 50000 bytes;

[0212] Maximum allowable delay deadline = 100 ms.

[0213] The information configured through the RRC reconfiguration message is as follows:

[0214] The trigger time threshold of DSR dsr_ThresholdList: the threshold corresponding to the 001 index is 25 ms;

[0215] The anti-jitter timing value dsr-ProhibitTimer: set to 10 ms.

[0216] The terminal device can set the PDCP discard timer to ∞, thereby disabling the PDCP discard timer.

[0217] After the terminal device receives the voice assistant task start request, it can create a context record AI_UplinkCtx:

[0218] The task identifier taskID of this task = 001;

[0219] The estimated number of bytes totalBytes of this task = 50000 bytes;

[0220] The maximum delay deadline allowed for this task = 100 ms.

[0221] After the voice task starts, the terminal device begins to send data. As the ACK feedback from the network side continuously arrives, the terminal device calculates the number of bytes that have been acknowledged. For example, the number of bytes that have been ACKed is currently 40000 bytes, so the remaining bytes Remain_Bytes = 50000 - 40000 = 10000 bytes. At the same time, the terminal device local timer T_elapsed records the time elapsed since the start of the task. Assuming the current time is 80 ms, the remaining time Remain_Time = deadline - T_elapsed = 100 ms - 80 ms = 20 ms.

[0222] Then, the terminal device can find the corresponding trigger threshold in dsr_ThresholdList in the trigger time threshold list of the DSR according to the first three indices of the task identifier taskID of this task, and the voice assistant task threshold is 25 ms. The current remaining time Remain_Time = 20 ms < 25 ms, and the time since the last DSR report has exceeded the anti-jitter time dsr-ProhibitTimer = 10 ms. Therefore, the terminal device will trigger the DSR.

[0223] The terminal device sends the DSR to the network device through the MAC CE. The DSR can carry the following information of the task:

[0224] The remaining time Remain_Time = 20 ms.

[0225] The remaining bytes Remain_Bytes = 10000 bytes.

[0226] The network device (such as a base station) parses the DSR MAC CE content, calculates the number of RB resources required to complete the transmission of 10000 bytes within 20 ms. Then, the network device immediately allocates the corresponding uplink grant to the terminal device to ensure timely transmission of the remaining data and avoid timeout.

[0227] Finally, the terminal device sends the remaining data segments using the uplink grant. If there is new data arriving and the initial uplink grant is not enough to cover, the terminal device initiates an additional request through the short-BSR-remaining field in the MAC PDU. The network device calculates additional resources according to the short-BSR-remaining information and issues an additional uplink grant.

[0228] It should be understood that Figures 1 to 9 The flowcharts or scenario charts shown are only for ease of understanding and are not intended to limit the embodiments of the present application to the examples shown in the charts. In fact, based on the examples in the above Figures 1 to 9 embodiments, equivalent transformations can be made to obtain more implementation manners.

[0229] Figure 10 is a schematic block diagram of a communication apparatus provided by an embodiment of the present application. The communication apparatus 1000 can be a chip, a chip system, or a processor, etc. of a terminal device for implementing the above method. The communication apparatus 1000 can be used to implement the method described in the above method embodiments, and specific implementation can be referred to the description in the above method embodiments.

[0230] As shown in Figure 10 The communication apparatus 1000 can include one or more processors 1001, which can also be referred to as processing units or processing modules, and can implement certain control functions. The processor 1001 can be a general-purpose processor or a special-purpose processor, etc., for example, a baseband processor or a central processing unit. The baseband processor can be used to process communication protocols and communication data, and the central processing unit can be used to control the communication apparatus 1000 (such as a base station, a baseband chip, a user, and a user chip), execute software programs, and process data of the software programs.

[0231] In a possible implementation manner, the processor 1001 can also store instructions and / or data, which can be executed by the processor 1001, so that the communication apparatus 1000 performs the method described in the above method embodiments.

[0232] In another possible implementation manner, the communication apparatus 1000 can include a communication interface 1002 for implementing receiving and sending functions. For example, the communication interface 1002 can be a transceiver circuit, an interface, an interface circuit, or a transceiver, etc. The transceiver circuit, the interface, the interface circuit, or the transceiver for implementing the receiving and sending functions can be separate or integrated together. The above transceiver circuit, interface, interface circuit, or transceiver can be used for reading and writing of codes / data, or the above transceiver circuit, interface, interface circuit, or transceiver can be used for transmission or transfer of signals.

[0233] Optionally, the communication apparatus 1000 can include one or more memories 1003 that can store instructions that can be executed by the processor 1001 to cause the communication apparatus 1000 to perform the methods described in the above method embodiments. Optionally, the memories 1003 can also store data. Optionally, the processor 1001 can also store instructions and / or data. The processor 1001 and the memories 1003 can be separately arranged, or can be integrated together.

[0234] It should be understood that, in a possible implementation, the steps in the method embodiments provided in the present application can be completed by integrated logic circuits of hardware in the processor or instructions in the form of software. The steps of the method disclosed in combination with the embodiments of the present application can be directly embodied as being completed by a hardware processor, or being completed by a combination of hardware and software modules in the processor. The software modules can be located in random access memories, flash memories, read-only memories, programmable read-only memories or electrically erasable programmable memories, registers, or other mature storage mediums in the art. The storage medium is located in the storage, and the processor reads information in the storage, and combines the hardware to complete the steps of the above method. To avoid repetition, it will not be described in detail here.

[0235] In one implementation, the communication apparatus 1000 can correspond to the terminal device in the above method embodiments, and can be used to execute the steps and / or processes performed by the terminal device in the above method embodiments. The processor 1001 can be used to execute the instructions stored in the memory 1003, and when the processor 1001 executes the instructions stored in the memory, the processor 1001 is used to execute the steps and / or processes of the above method embodiments corresponding to the terminal device.

[0236] In another implementation, the communication apparatus 1000 can correspond to the network device in the above method embodiments, and can be used to execute the steps and / or processes performed by the network device in the above method embodiments. The processor 1001 can be used to execute the instructions stored in the memory 1003, and when the processor 1001 executes the instructions stored in the memory, the processor 1001 is used to execute the steps and / or processes of the above method embodiments corresponding to the network device.

[0237] It should be appreciated that the above-mentioned processor can be one or more chips. For example, the processor can be a field programmable gate array (FPGA), can be an application specific integrated circuit (ASIC), can also be a SoC, can also be a central processor unit (CPU), can also be a network processor (NP), can also be a digital signal processing circuit (digital signal processor, DSP), can also be a micro controller unit (MCU), can also be a programmable logic device (programmable logic device, PLD) or other integrated chip.

[0238] It can be understood that the memory in the embodiments of the present application can be a volatile memory or a non-volatile memory, or can include both volatile and non-volatile memories. Among them, the non-volatile memory can be a read-only memory (read-only memory, ROM), a programmable read-only memory (programmable ROM, PROM), an erasable programmable read-only memory (erasable PROM, EPROM), an electrically erasable programmable read-only memory (electrically EPROM, EEPROM) or a flash memory. The volatile memory can be a random access memory (random access memory, RAM) used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as static random access memory (static RAM, SRAM), dynamic random access memory (dynamic RAM, DRAM), synchronous dynamic random access memory (synchronous DRAM, SDRAM), double data rate synchronous dynamic random access memory (double data rate SDRAM, DDR SDRAM), enhanced synchronous dynamic random access memory (enhanced SDRAM, ESDRAM), synchronous link dynamic random access memory (synchlink DRAM, SLDRAM) and direct memory bus random access memory (direct rambus RAM, DR RAM). It should be noted that the memory of the system and method described herein is intended to include, but not limited to, these and any other suitable types of memory.

[0239] According to the method provided in the embodiments of the present application, the present application further provides a chip system, which comprises one or more processors, and is used for calling and running instructions stored in a memory, so that the method provided in the embodiments of the present application is executed. The chip system can be composed of a chip, or can comprise a chip and other discrete devices. The chip system can comprise an input circuit or an interface for transmitting information or data, and an output circuit or an interface for receiving information or data.

[0240] According to the method provided in the embodiments of the present application, the present application further provides a computer program product, which comprises computer program codes, and when the computer program codes are run on a computer, the computer is caused to execute the steps or processes in any of the foregoing method embodiments.

[0241] According to the method provided in the embodiments of the present application, the present application further provides a computer readable storage medium, which stores program codes, and when the program codes are run on a computer, the computer is caused to execute the steps or processes in any of the foregoing method embodiments.

[0242] The computer readable storage medium can be the volatile memory or the non-volatile memory described above, or can comprise the volatile memory and the non-volatile memory at the same time.

[0243] In the embodiments of the present application, each term and English abbreviation is an exemplary example given for convenience of description, and should not constitute any limitation on the present application. The present application does not exclude the possibility of defining other terms capable of achieving the same or similar functions in the existing or future protocols.

[0244] In the foregoing embodiments, all or part of the embodiments can be realized by software, hardware, firmware or any combination thereof. When realized by software, all or part of the embodiments can be realized in the form of a computer program product. The computer program product comprises one or more computer instructions. When the computer instructions are loaded and executed on a computer, all or part of the processes or functions according to the embodiments of the present application are generated.

[0245] In the several embodiments provided in the present application, it should be understood that the disclosed system, device and method can be implemented in other manners. For example, the described device embodiments are merely schematic. The division of the units is merely a logical function division. There can be another division manner for the actual implementation, for example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed mutual couplings or direct couplings or communication connections between the units can be indirect couplings or communication connections through some interfaces, devices or units, and can be electrical, mechanical or in other forms.

[0246] It should be understood that the size of the serial number of each process in various embodiments of the present application does not mean the order of execution, the execution order of each process should be determined by its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present application. In summary, the above is only a preferred embodiment of the technical scheme of the present application, and is not used to limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A DSR transmission method, characterized by, The method is applied to a terminal device, and the method comprises: obtaining a data packet of a first AI non-video task, the first AI non-video task being an AI non-video task based on discrete events or instructions and relying on AI technology to complete inference on a cloud device; starting to send the data packet of the first AI non-video task to the cloud device through a network device; in a case where a remaining duration of the first AI non-video task is less than or equal to a first trigger time threshold, sending a DSR to the network device, the DSR comprising a remaining data amount, the remaining data amount being an amount of data of data packets estimated to be transmitted in the first AI non-video task, the remaining data amount being used to trigger the network device to reserve resources, the remaining duration being determined according to a maximum delay allowed by the first AI non-video task and a time duration consumed from starting to send the data packet of the first AI non-video task to an update of a sending window; receiving a first uplink grant from the network device; continuing to send the data packets to be transmitted in the first AI non-video task to the cloud device through the network device on resources indicated by the first uplink grant.

2. The method of claim 1, wherein, The first AI non-video task belongs to a first type of AI non-video task, and different types of AI non-video tasks correspond to different trigger time thresholds.

3. The method of claim 1, wherein, Before starting to send the data packet of the first AI non-video task to the cloud device through the network device, the method further comprises: setting a duration of a PDCP discard timer of the first AI non-video task to infinity.

4. The method of claim 1, wherein, The sending of the DSR to the network device in the case where the remaining duration of the first AI non-video task is less than or equal to the first trigger time threshold comprises: each time the sending window of the first AI non-video task is updated, obtaining the remaining data amount and the remaining duration of the first AI non-video task; in a case where the remaining duration of the first AI non-video task is less than or equal to the first trigger time threshold and a first duration is greater than or equal to an anti-jitter timing value, sending the DSR to the network device, the first duration being a time duration from a last time the DSR is sent to the network device to the update of the sending window.

5. The method of claim 4, wherein, The obtaining of the remaining data amount and the remaining duration of the first AI non-video task comprises: determining the remaining data amount according to an estimated data amount of the first AI non-video task and a data amount that has been confirmed by the cloud device; determining the remaining duration of the first AI non-video task according to the maximum delay allowed by the first AI non-video task and the time duration consumed from starting to send the data packet of the first AI non-video task to the update of the sending window.

6. The method of claim 4, wherein, The method further comprises: receiving a radio resource control reconfiguration message from the network device; wherein the radio resource control reconfiguration message comprises the anti-jitter timing value and trigger time thresholds corresponding to a plurality of AI non-video tasks, and different types of AI non-video tasks correspond to different trigger time thresholds.

7. The method according to any one of claims 1 to 6, characterized in that, The DSR further comprises a remaining duration of the first AI non-video task, and the remaining duration of the first AI non-video task is used together with the remaining data amount to trigger the network device to reserve resources.

8. The method according to any one of claims 1 to 6, characterized in that, The method further comprises: In a case where a number of resources required by the data packets to be transmitted in the first AI non-video task exceeds the number of resources indicated by the first uplink grant, sending a short buffer status report to the network device, the short buffer status report comprising a first index value, the first index value being used to trigger the network device to additionally reserve resources; receiving a second uplink grant from the network device; continuing to send the data packets to be transmitted in the first AI non-video task to the cloud device through the network device on the resources additionally reserved by the second uplink grant.

9. The method of claim 8, wherein, The buffer of the short buffer status report is composed of P bits, and Q bits of the P bits are used to write an index value, and different index values are used to additionally reserve different numbers of resources, and P and Q are positive integers.

10. The method of claim 8, wherein, The sending of the short buffer status report to the network device comprises: sending a MAC PDU to the network device, the MAC PDU comprising a first subheader, a second subheader, the short buffer status report, and a MAC SDU next to the short buffer status report, a logical channel identifier of the first subheader pointing to the short buffer status report, a logical channel identifier of the second subheader pointing to the MAC SDU, and the MAC SDU comprising a next data packet to be sent.

11. The method of claim 8, wherein, The first uplink grant and the second uplink grant have the same hybrid automatic repeat request process identifier.

12. A DSR transmission method, characterized by, The method is applied to a network device, and the method comprises: starting to receive data packets of a first AI non-video task from a terminal device and forwarding the data packets of the first AI non-video task to a cloud device, the first AI non-video task being an AI non-video task generated based on discrete events or instructions and relying on AI technology to complete inference at the cloud device; receiving a DSR from the terminal device, the DSR comprising a remaining data amount, the remaining data amount being a data amount of data packets estimated to be transmitted in the first AI non-video task, the DSR being sent by the terminal device in a case where a remaining duration of the first AI non-video task is less than or equal to a first trigger time threshold, the remaining duration being determined according to a maximum delay allowed by the first AI non-video task and a time duration consumed from the terminal device starting to send the data packets of the first AI non-video task to a sending window being updated; reserving resources for the first AI non-video task according to the remaining data amount; sending a first uplink grant to the terminal device, the first uplink grant being used to indicate the resources reserved for the first AI non-video task; continuing to receive the data packets of the first AI non-video task from the terminal device and continuing to forward the data packets of the first AI non-video task to the cloud device on the resources indicated by the first uplink grant.

13. The method of claim 12, wherein, The reserving resources for the first AI non-video task according to the residual data amount comprises: determining the number of resources reserved for the first AI non-video task according to the residual data amount and the number of bits contained in a transport block.

14. The method of claim 12, wherein, The DSR further comprises a residual time length of the first AI non-video task; and the reserving resources for the first AI non-video task according to the residual data amount comprises: determining a minimum instantaneous rate according to the residual data amount and the residual time length of the first AI non-video task; determining the number of bits contained in a transport block according to the minimum instantaneous rate, different minimum instantaneous rates corresponding to different numbers of bits; determining the number of resources reserved for the first AI non-video task according to the residual data amount and the number of bits contained in a transport block.

15. The method according to any one of claims 12 to 14, characterized in that, After the sending of the first uplink grant to the terminal device, the method further comprises: receiving a short buffer status report from the terminal device, the short buffer status report comprising a first index value; appending resources for the first AI non-video task according to the first index value; sending a second uplink grant to the terminal device; continuing to receive data packets of the first AI non-video task from the terminal device on resources indicated by the second uplink grant and continuing to forward the data packets of the first AI non-video task to the cloud device.

16. The method of claim 15, wherein, The buffer of the short buffer status report is composed of P bits, Q bits of the P bits being used to write index values, different index values being used to append different numbers of resources, P and Q being positive integers.

17. The method of any one of claims 12-14, wherein, The method further comprises: sending a radio resource control reconfiguration message to the terminal device; wherein the radio resource control reconfiguration message comprises a jitter prevention timing value and trigger time thresholds corresponding to a plurality of AI non-video tasks, different types of AI non-video tasks corresponding to different trigger time thresholds.

18. A terminal device, comprising: The terminal device comprises at least one processor coupled with a memory, the memory storing programs or instructions, the processor executing the programs or instructions to enable the terminal device to perform the method of any one of claims 1 to 11.

19. A network device, comprising: The network device comprises at least one processor coupled with a memory, the memory storing programs or instructions, the processor executing the programs or instructions to enable the network device to perform the method of any one of claims 12 to 17.

20. A communication system, characterized by The communication system comprises a terminal device, a network device and a cloud device; wherein the terminal device is configured to perform the method of any one of claims 1 to 11, the network device is configured to perform the method of any one of claims 12 to 17, and the cloud device is configured to complete inference relying on AI technology.

21. A computer-readable storage medium, characterized in that, The computer readable storage medium stores computer programs or instructions which, when executed, enable a computer to perform the method of any one of claims 1 to 17.

Images