Device and method for managing a set of positioning models, in particular artificial intelligence or machine learning positioning models

Abstract

A device (102) and a method for managing a set of positioning models, in particular artificial intelligence or machine learning positioning models, wherein the method comprises sending a deactivation signal (206) to a first positioning model (202) of the set of positioning models, in particular sending a first message comprising the deactivation signal (206), and sending an activation signal (208) to a second positioning model (204) of the set of positioning models, in particular sending a second message comprising the activation signal (208). The invention further relates to a method for operating the first positioning model (202).

Description

background

[0001] The invention relates to a device and a method for managing a set of positioning models, in particular artificial intelligence or machine learning positioning models.

[0002] AI (artificial intelligence) or ML (machine learning)-based positioning systems in wireless communication networks require the deployment and management of AI / ML positioning models, which necessitates efficient lifecycle management (LCM). Effective LCM is crucial for maintaining the accuracy and reliability of the AI / ML positioning models, especially in dynamic environments where conditions can change rapidly and unpredictably. Disclosure of the invention

[0003] The device and method according to the independent claims provide a framework for managing a set of positioning models, in particular artificial intelligence or machine learning positioning models, in a wireless communication network, particularly depending on their performance.

[0004] A method for managing a set of positioning models, in particular artificial intelligence (AI) or machine learning (ML) positioning models, in a wireless communication network, wherein the method comprises sending a deactivation signal to a first positioning model of the set of positioning models, in particular sending a first message containing the deactivation signal, and sending an activation signal to a second positioning model of the set of positioning models, in particular sending a second message containing the activation signal. This framework enables switching between the positioning models, i.e., deactivating the first positioning model and activating the second positioning model, thereby ensuring a continuous positioning service.The procedure can be part of a Location Management Function (LMF).

[0005] The procedure can include determining the performance of the first positioning model and sending a deactivation signal to the first positioning model and an activation signal to the second positioning model if its performance falls below a performance threshold. This means that switching from the first to the second positioning model is triggered by switching signals if the performance of the first positioning model has fallen below a predefined threshold.

[0006] The procedure can include determining performance based on the accuracy, latency, and / or resource consumption of the first positioning model. This means that switching from the first to the second positioning model is based on accuracy, latency, or resource consumption.

[0007] The procedure may involve receiving the accuracy and / or latency and / or resource consumption of the first positioning model from the first positioning model, in particular receiving a third message containing the accuracy, latency, or resource consumption of the first positioning model. This means that the first positioning model reports its status with respect to accuracy, latency, or resource consumption.

[0008] The method can include sending an instruction to adapt at least one model parameter of the second positioning model to the second positioning model, in particular sending a fourth message containing the instruction. The instruction can be sent based on the determined performance of the first positioning model, preferably if the performance is below the performance threshold. The at least one model parameter includes, for example, a configuration parameter for initializing the second positioning model. This makes it possible to activate the second positioning model in an initialized state.

[0009] A method for operating the first positioning model, in particular an artificial intelligence or machine learning positioning model, in the wireless communication network comprises receiving a deactivation signal, in particular receiving a first message containing the deactivation signal and deactivating the first positioning model upon receipt of the deactivation signal, or receiving an activation signal, in particular receiving a second message containing the activation signal and activating the first positioning model upon receipt of the activation signal. This seamlessly activates or deactivates the first positioning model.

[0010] The procedure for operating the first positioning model may include determining and transmitting an accuracy and / or latency and / or resource consumption value of the first positioning model, in particular transmitting a third message containing the accuracy, latency, or resource consumption of the first positioning model. This reports the status of the first positioning model.

[0011] The procedure for operating the first positioning model can include receiving an instruction to adjust at least one model parameter of the first positioning model, in particular receiving a fourth message containing the instruction, and adjusting the at least one model parameter according to the instruction. This adjusts the first positioning model accordingly.

[0012] The procedure for operating the first positioning model can include determining the performance of the first positioning model and sending a switching signal to a second positioning model, in particular sending a fifth message containing the switching signal if the performance falls below a performance threshold. This switching signal enables a seamless transition from the first positioning model to another positioning model.

[0013] The procedure for operating the first positioning model can include determining the accuracy of the first positioning model and determining its performance as a function of this accuracy. Additionally or alternatively, the procedure can include determining the latency of the first positioning model and determining its performance as a function of this latency. Additionally or alternatively, the procedure can include determining the resource consumption of the first positioning model and determining its performance as a function of this resource consumption.

[0014] The method for operating the first positioning model may include receiving a switching signal, in particular receiving a fifth message containing the switching signal, and activating the positioning model upon receiving the switching signal. This switching signal enables a seamless transition from another positioning model to the first positioning model.

[0015] A device for managing a set of positioning models, in particular artificial intelligence or machine learning positioning models, or for operating a first positioning model, in particular an artificial intelligence or machine learning positioning model, in a wireless communication network, is configured to perform the method(s). The device is, for example, a network node that provides the location management function (LMF).

[0016] A computer program may be provided, wherein the computer program includes computer-readable instructions which, when executed by a computer and / or device according to an embodiment, cause the computer and / or device to perform the procedure.

[0017] Some embodiments relate to a computer-readable storage medium comprising instructions which, when executed by a computer and / or the device, cause the computer and / or the device to carry out the method according to the disclosure.

[0018] Some embodiments relate to a data carrier signal that carries and / or characterizes the computer program according to the disclosure.

[0019] Further embodiments are described below and shown in the drawing. The drawing shows: Fig. 1. Schematic representation of a wireless communication network, Fig. 2 messages regarding the management of a set of positioning models.

[0020] Fig. Figure 1 schematically shows a wireless communication network 100. The wireless communication network 100 is, for example, a cellular communication network 100.

[0021] The wireless communication network 100 can be based on at least one radio standard of the third generation partnership project, 3GPP, and / or at least partially conform to such a standard, for example 4G (fourth generation), 5G (fifth generation) or 6G (sixth generation), or another radio access technology.

[0022] The wireless communication network 100 includes a device 102, for example a network node, which is configured to provide and manage a set of positioning models. The device 102 can provide LMF.

[0023] Fig.Item 2 displays messages related to managing the set of positioning models. These positioning models include, for example, positioning models based on artificial intelligence (AI) or machine learning (ML).

[0024] The wireless communication network 100 can include a positioning model repository 200 which is configured to provide the positioning models for the communication network 100, in particular via the device 102.

[0025] A procedure for managing the set of positioning models is described for a first positioning model 202 and a second positioning model 204 of the set of positioning models. The procedure can coincide with or be part of the Position Management Function (LMF) of Communication Network 100. Alternatively, the procedure can be executed as a separate network function within Communication Network 100. An instance of the LMF or the separate network function can be associated with a Network Function Instance Identification (NF instance ID).

[0026] The procedure is not limited to managing two positioning models. It can be applied to more than two positioning models from the set of positioning models, as described using the two example positioning models.

[0027] The process may include automated model provisioning, which may include, for example, the following automated processes: Deployment automation: Provisioning automation includes, for example, the use of 3GPP-defined Configuration and Provisioning of Network Functions (CPNF) messages to set up the AI / ML positioning models. This ensures that the AI / ML positioning models are automatically configured, instantiated, and validated, for example, according to the 3GPP SA5 specifications.

[0028] Following an example, the procedure uses a deployment script. This script utilizes, for instance, the 3GPP-NFV-MANO architectural frameworks to orchestrate the instantiation of AI / ML positioning models via Virtual Network Functions (VNFs) and Cloud-native Network Functions (CNFs), as well as lifecycle management. The scripts employ standardized 3GPP interfaces for verifying and validating model deployment. Dynamic model activation and deactivation:

[0029] The procedure includes sending a deactivation signal 206 to the first positioning model 202.

[0030] The method includes, for example, sending a first message containing the deactivation signal 206. The method further includes sending an activation signal 208 to the second positioning model 204. The method also includes, for example, sending a second message containing the activation signal 208.

[0031] The deactivation signal 206 can include a first model identifier of the first positioning model 202. This means that the deactivation signal 206 includes the identifier of the model to be deactivated.

[0032] The first message may be a Model Function Deactivation (MFD) message.

[0033] The first message can include a first field containing the NF instance identifier.

[0034] The first message may include a second field containing a reason for deactivation.

[0035] The first message is triggered, for example, by a network condition that is detected, in particular, by monitoring a network function (NF monitoring). The first message is triggered, for example, by a network condition that no longer justifies the use of a particular AI / ML positioning model for improved positional accuracy.

[0036] The activation signal 208 can include a second model identifier of the second positioning model 204. This means that the activation signal 208 includes the identifier of the model to be activated.

[0037] The second message can be a Model Function Activation (MFA) message.

[0038] The second message can include a first field containing the NF instance identifier.

[0039] The second positioning model 204 can include configurable parameters. The second message can include a second field containing the configuration parameters to be adjusted.

[0040] The second positioning model 204 can be configured to be activated when a trigger condition is met. The second message can include a third field containing the trigger condition.

[0041] The second message is triggered, for example, by a network condition that is detected, in particular, by monitoring a network function. The second message is triggered, for example, by a network condition that requires the activation of a specific AI / ML positioning model for improved positioning accuracy.

[0042] According to the exemplary embodiment, the device 102 provides the first positioning model 202 and the second positioning model 204. This means that the deactivation signal 206 and the activation signal 208 are transmitted within the device 102.

[0043] The method is not limited to transmitting the deactivation signal 206 or the activation signal 208 within the device 102. The deactivation signal 206 can be sent to a network node of the wireless communication network 100 that provides the second positioning model 204. The activation signal 208 can be sent to a network node of the wireless communication network 100 that provides the first positioning model 202. The same network node can provide both the first positioning model 202 and the second positioning model 204. The deactivation signal 206 and the activation signal 208 can be sent to the same network node. This framework allows the network node(s) to deactivate the first positioning model 202 and activate the second positioning model 204, thereby ensuring a continuous positioning service.

[0044] The procedure may include determining the performance of the first positioning model 202 and sending the deactivation signal 206 to the first positioning model 202 and sending the activation signal 208 to the second positioning model 204 if the performance is below a performance threshold.

[0045] This means that switching from the first positioning model 202 to the second positioning model 204 is triggered by switching signals when the performance of the first positioning model 202 falls below a predefined threshold. The second positioning model 204 is activated, for example, as a fallback model for the first positioning model 202.

[0046] The procedure continuously monitors environmental conditions, such as signal strength and interference levels in the communication network. If the conditions indicate that a specific positioning model should be activated (e.g., in the case of strong multipath interference), the procedure involves sending the activation signal 208 containing the model identifier and configuration parameters. Conversely, if the conditions no longer require the use of the positioning model, the deactivation signal 206 containing the model identifier is sent to shut down the positioning model in a controlled manner.

[0047] The method can include determining performance based on the accuracy, latency, or resource consumption of the first positioning model. This means that switching from the first to the second positioning model is based on accuracy, latency, or resource consumption.

[0048] The procedure may involve receiving the accuracy, latency, or resource consumption of the first positioning model 202 from the first positioning model 202, in particular receiving a third message 210 containing the accuracy, latency, or resource consumption of the first positioning model 202. This means that the first positioning model 202 reports its status with respect to accuracy, latency, or resource consumption.

[0049] The procedure includes determining and transmitting the accuracy, latency, or resource consumption of the first positioning model 202, in particular through the third message.

[0050] The procedure may include receiving the accuracy, latency, or resource consumption of the second positioning model 204 from the second positioning model 204, in particular receiving the third message 210, which includes the accuracy, latency, or resource consumption of the second positioning model 204. This means that the second positioning model 204 reports its status with respect to accuracy, latency, or resource consumption.

[0051] The procedure can include determining the accuracy of the first positioning model 202 and determining the performance as a function of the accuracy.

[0052] The procedure can include determining the latency of the first positioning model 202 and determining the performance of the first positioning model 202 as a function of the latency.

[0053] The procedure can include determining the resource consumption of the first positioning model 202 and determining the performance of the first positioning model 202 as a function of resource consumption.

[0054] Performance assessment can integrate management services defined in 3GPP-TS-28.552 that monitor the performance of network functions and AI / ML positioning models, providing real-time data for operational decision-making. Performance assessment can utilize telemetry data aligned with 3GPP-defined performance management (PM) services. Performance assessment can leverage collected data to perform trend analysis and anomaly detection, and to initiate preventive or corrective actions in accordance with 3GPP Management and Orchestration (MANO) standards.

[0055] For example, if the performance of an AI / ML positioning model falls below the defined Quality of Service (QoS) parameters of communication network 100, a Session and Service Continuity (SSC) mode is triggered to switch to the fallback model. The second positioning model 204 can be selected as the fallback model from the set of positioning models, depending on its ability to achieve higher accuracy than the first positioning model 202.

[0056] The procedure may include sending an instruction 212 to adapt at least one model parameter of the second positioning model 204 to the second positioning model 204.

[0057] The procedure may include sending a fourth message containing instruction 212.

[0058] The at least one model parameter includes, for example, a configuration parameter for initializing the second positioning model 204. This allows the network node providing the second positioning model 204 to activate the second positioning model 204 in its initialized state.

[0059] The fourth message can be a Model Adjustment Instruction (MAI). The fourth message can include a first field for identifying a target positioning model. This first field might contain, for example, the second model identifier.

[0060] The fourth message can include a second field to specify the adjustment of at least one model parameter. This second field might include, for example, the adjustment.

[0061] The fourth message can include a third field for specifying a reason for the adjustment. This third field could, for example, contain the reason for the adjustment.

[0062] Sending the adjustments enables dynamic adaptation of the AI / ML positioning model that receives the adjustment. This adaptation can be based on real-time analysis, thus ensuring optimal performance and compliance with 3GPP QoS requirements.

[0063] The procedure includes receiving the instruction, in particular the fourth message, and adjusting at least one parameter in the second positioning model 204.

[0064] The procedure includes receiving the deactivation signal 206, for example the first message containing the deactivation signal 206, and deactivating the first positioning model 202 upon receiving the deactivation signal 206.

[0065] The procedure includes receiving the activation signal 208, for example the second message containing the activation signal 208, and activating the second positioning model 204 upon receiving the activation signal 208.

[0066] The procedure may include determining and sending the accuracy, latency, or resource consumption of the positioning models from the set of positioning models.

[0067] The procedure may include transferring the third message from the positioning models in the set of positioning models.

[0068] The method for operating the first positioning model 202 may include sending a switching signal 214, in particular sending a fifth message comprising the switching signal 214, to the second positioning model 204 if the performance of the first positioning model 202 is below a performance threshold.

[0069] The fifth message can include a first field to identify the current positioning model. This first field might contain, for example, the first model identifier. The fifth message can include a second field to identify the target positioning model. This second field might contain, for example, the second model identifier. The fifth message can include a third field to specify a reason for switching. The reason for switching might be, for example, insufficient accuracy, excessive latency, or excessive resource consumption.

[0070] The switching signal 214 enables a seamless transition from the first positioning model 202 to the second positioning model 204, thus ensuring an uninterrupted positioning service and compliance with the 3GPP-SSC modes.

[0071] The method for operating the second positioning model 204 may include receiving the switching signal 214, in particular receiving the fifth message which includes the switching signal 214, and activating the second positioning model 204 upon receiving the switching signal 214.

[0072] The procedure may include the operation of the positioning models from the set of positioning models as described for the first positioning model 202 and / or the second positioning model 204.

Claims

[1] Methods for managing a set of positioning models, in particular artificial intelligence or machine learning positioning models, characterized by , that the procedure includes: - Sending a deactivation signal (206) to a first positioning model (202) of the set of positioning models, in particular sending a first message comprising the deactivation signal (206), and - Sending an activation signal (208) to a second positioning model (204) of the set of positioning models, in particular sending a second message that includes the activation signal (208). [2] Method according to claim 1, characterized by , that the procedure includes: - Determining the performance of the first positioning model (202), and - Sending the deactivation signal (206) to the first positioning model (202) and the activation signal (208) to the second positioning model (204) when the performance is below a performance threshold. [3] Method according to claim 2, characterized by , that the procedure includes: - Determining the performance depending on an accuracy and / or a latency and / or a resource consumption of the first positioning model (202). [4] Method according to claim 3, characterized by , that the procedure includes: - Receiving the accuracy and / or latency and / or resource consumption of the first positioning model (202) from the first positioning model (202), in particular receiving a third message that includes the accuracy, latency or resource consumption of the first positioning model (202). [5] Method according to any one of the preceding claims, characterized by , that the procedure includes: - Sending an instruction to adapt at least one model parameter of the second positioning model (204) to the second positioning model (204), in particular sending a fourth message comprising the instruction. [6] Method for operating a first positioning model (202), in particular an artificial intelligence or machine learning positioning model, in a wireless communication network, characterized by , that the procedure includes: - Receiving a deactivation signal (206), in particular receiving an initial message comprising the deactivation signal (206), and - Deactivating the first positioning model (202) upon receiving the deactivation signal (206), or - Receiving an activation signal (208), in particular receiving a second message comprising the activation signal (208), and - Activating the first positioning model (202) upon receipt of the activation signal (208). [7] Method according to claim 6, characterized by , that the procedure includes: - Determining and transmitting an accuracy and / or a latency and / or a resource consumption of the first positioning model (202), in particular transmitting a third message that includes the accuracy, the latency or the resource consumption of the first positioning model (202). [8] Method according to one of claims 6 or 7, characterized by , that the procedure includes: - Receiving an instruction to adjust at least one model parameter of the first positioning model (202), in particular receiving a fourth message comprising the instruction, and - Adjusting at least one model parameter according to the instructions. [9] Method according to any one of claims 6 to 8, characterized by , that the procedure includes: - Determining the performance of the first positioning model (202), and - Sending a switching signal to a second positioning model (204), in particular sending a fifth message including the switching signal when the performance is below a performance threshold. [10] Method according to claim 9, characterized by , that the procedure includes: - Determining an accuracy of the first positioning model (202), and - Determining performance depending on accuracy, and / or - Determining a latency of the first positioning model (202), and - Determining performance depending on latency and / or - Determining the resource consumption of the first positioning model (202), and - Determining the performance of the positioning model depending on resource consumption. [11] Method according to any one of claims 6 to 10, characterized by , that the procedure includes: - Receiving a switching signal, in particular receiving a fifth message that includes the switching signal, and - Activating the positioning model upon receiving the switching signal. [12] Device (102) for managing a set of positioning models, in particular artificial intelligence or machine learning positioning models, or for operating a first positioning model (202), in particular an artificial intelligence or machine learning positioning model, in a wireless communication network, characterized by that the device is designed to carry out the method according to one of the preceding claims. [13] Computer program, characterized bythat the computer program comprises computer-readable instructions which, when executed by a computer and / or the device (102) according to claim 12, cause the computer and / or the device (102) to carry out the method according to any one of claims 1 to 11. [14] Computer-readable storage medium comprising instructions which, when executed by a computer and / or the device (102) according to claim 12, cause the computer and / or the device (102) to carry out the method according to any one of claims 1 to 11. [15] Data carrier signal that carries and / or characterizes the computer program according to claim 13.

Images