System for predicting a 5G user plane using control plane features and Granger causality for feature selection

Abstract

A system for predicting a user plane using control plane features and Granger causality includes a host device, a cloud computing server, and controllers. The controllers execute control logic, including a prediction application (PA), that receives sensor data from the sensors, sends control and user plane data to an infrastructure and to the cloud computing servers, and accesses prior knowledge data stored in the memory of the cloud computing servers. Additional control logic performs Granger causality testing on the user and control plane data and uses a prediction model to generate a prediction from fused user and control plane data.A prediction checker is applied to the prediction from the prediction model and the PA enables the host device to make reliable predictions that allow the host device to adapt to conditions of dynamic wireless communication networks even when a best-fit model is not fully trained.

Description

INITIATIONThe present invention relates to cellular communication systems and, more particularly, to a system for predicting a user level using control level features and Granger causality.DE 10 2021 214 675 A1 discloses a method for predicting quality parameters of a communication network, mobile components such as vehicles being supplied. With the aid of a neural network, local variables are predicted on the basis of measured values.Further prior art is evident from the publications U.S. Pat. No. 2019) / 0 319 868 A1 and WO 2022 / 144 582 A1.Cellular networks are increasingly being used by a variety of devices including vehicles. Vehicles often use cellular networks to communicate with other vehicles, infrastructure, and the like. To optimize performance, manage network traffic, and improve user experience, machine learning techniques are often used for feature selection. However, conventional machine learning techniques often do not capture relationships between control plane information and user plane metrics in dynamic environments.Accordingly, while current systems and methods for optimizing performance, managing network traffic, and improving user experience through user plane metric predictions achieve their intended purpose, there is a need for new and improved systems and methods for predicting user plane metrics using control plane features that are highly adjustable, work well in dynamic environments, improve mobile communication reliability and user experience, and can be easily applied to new and existing platforms without increasing system complexity.SUMMARYAccording to the invention there is described a system for predicting a user level using control level features and Granger causality, characterised by the features of claim 1.The system includes: a host device having one or more sensors. The one or more sensors detect telematics information, telecommunication information, host device stereometry information, and host device position information. The system further includes one or more cloud computing servers and one or more controllers. Each of the one or more controllers includes a processor, a memory, and one or more input / output (I / O) ports. The input / output ports are in communication with the one or more sensors and the one or more cloud computing servers. The memory stores programmatic control logic. The processor executes the programmatic control logic. The programmatic control logic contains a prediction application (PA). The PA includes at least first, second, third, fourth, fifth and sixth control logic portions. The first control logic receives sensor data from the one or more sensors. The second control logic sends control plane data and user plane data to an infrastructure and to the cloud computing servers via the input / output ports. The third control logic accesses prior knowledge data stored in the storage of the cloud computing server. The fourth control logic performs a Granger causality check on the user plane data and the control plane data. The fifth control logic uses a prediction model to generate a prediction from merged user plane data and control plane data. The sixth control logic applies a prediction checker to the prediction from the prediction model. The PA allows the host device to make reliable predictions that allow the host device to adapt to conditions of dynamic wireless communication networks even if a best possible model is not fully trained.In another aspect of the present invention, the host device further defines a vehicle that communicates with the cloud computing servers via a wireless communication network. The first control logic further includes obtaining telematics information, telecommunication information, vehicle telemetry information, and vehicle position information from the one or more sensors.In yet another aspect of the present invention, the second control logic further includes control logic for sending control plane data radio resource control (RRC) messages to an infrastructure including one or more cell towers and to one or more cloud computing servers. The second control logic also receives control plane data RRC messages from the one or more cell towers and from the one or more cloud computing servers and sends user plane data to the one or more cell towers and to the one or more cloud computing servers. The second control logic further receives user plane data from the one or more cell towers and from the one or more cloud computing servers.In yet another aspect of the present invention, the third control logic further includes control logic for accessing prior knowledge data including data obtained from sensors of the host device and from additional devices, data obtained from global positioning system (GPS) satellites, and data obtained from an infrastructure including one or more cell towers. The third control logic further compares data control plane data RRC messages from the host device with the prior knowledge data to generate a partial input to the prediction checker.In yet another aspect of the present invention, the fourth control logic further includes control logic for applying a Granger causality check to the control plane data RRC messages and to the user plane data. The Granger causality check outputs a P value that predicts when a change in the control plane data RRC messages caused a change in the user plane data. The Granger causality check is applied using at least three different P-value based selections including: a threshold P-value selection, a fixed number P-value selection, and a conditional selection. In the threshold P-value selection, all control plane data RRC features having a P-value less than a predetermined threshold are used. In the P-value selection by a fixed number, control plane data RRC features are classified from smallest to largest, and a predetermined number of features having the smallest P-values are selected. In the conditional selection, control plane data RRC features from the P-value selection are compared by threshold and from the P-value selection by a fixed number. When it is determined that the P-value selection has generated a P-value less than the P-value associated with the threshold P-value selection by a predetermined number, the P-value selection indicates a Granger causality between the control plane data RRC features and the user plane data by a predetermined number. When it is determined that the threshold P-value selection has generated a P-value that is less than the P-value associated with the threshold P-value selection by a predetermined number, the threshold P-value selection indicates a Granger causality between the control plane data RRC features and the user plane data. When it is determined that the threshold P-value selection and the fixed number P-value selection values are equal to each other, the threshold P-value selection and the fixed number P-value selection similarly indicate a Granger causality between the control plane data RRC features and the user plane data.In yet another aspect of the present invention, the fifth control logic further includes control logic for using the output of the Granger causality check to reduce an amount of data input to the prediction model from a first amount of data to a second amount of data less than the first and preventing overfeeding of the prediction model. The fifth control logic also uses a long short-term memory (LSTM) model and / or recurrent neural network (RNN) model and / or autoregressive integrated sliding average model (ARIMA) to generate a multivariate prediction from user plane data and control plane data output from the Granger causality check.In yet another aspect of the present invention, the fifth control logic further includes control logic for online training the PA by: initializing the PA as a null feature model as a reference, sending control plane messages including Granger causality check result information and a prediction error effective value (RMSE) of a current vehicle-side null feature model to the cloud computing server, evaluating a vehicle-side null feature model prediction RMSE training speed, and predicting and optionally providing a new model to the host device.In yet another aspect of the present invention, the sixth control logic further includes control logic for receiving data indicating that a trigger event has occurred and initializing the predictive checker with prior knowledge and having a user plane time series u(n) and a control plane time series C(n) as inputs. The sixth control logic further generates a weighted fused statistical prediction according to: p f= w1·p s+ w2·p l, where w1 is a first weight applied to a statistical prediction p s generated by a statistical prediction model as applied to the prior knowledge and the user plane time series, w2 is a second weight different from the first weight and applied to an LSTM prediction p l and p f is a fused prediction applied to the fused user plane and control plane data.In yet another aspect of the present invention, the control logic of the PA further includes: control logic that applies non-uniform upsampling in a data collection and time-series formulation for both the user-level and control-level data.In yet another aspect of the present invention, non-uniform upsampling further includes control logic that reduces bandwidth consumption and processing complexity from a first level to a second level that is less than the first level by obtaining user-level and control-level data only due to the occurrence of a triggering event and at one or more additional times after a predefined time δt after the triggering event. When a value of the control plane time series at time δt after the triggering event is greater than or equal to a value of the control plane time series at a subsequent triggering event, non-uniform upsampling causes the PA to increase the control plane time series to collect data due to the occurrence of a subsequent triggering event, and when a value of the control plane time series at time δt after the triggering event is less than the value of the control plane time series at the subsequent triggering event causes the PA to upsampling the control plane time series.Further described is a method for predicting a user level using control level features and Granger causality, comprising: detecting telematics information, telecommunication information, host device stereometry information, and host device position information with one or more sensors mounted on a host device, using one or more cloud computing servers, and using one or more controllers. Each of the one or more controllers includes a processor, a memory, and one or more input / output (I / O) ports. The input / output ports are in communication with the one or more sensors and the one or more cloud computing servers. The memory stores programmatic control logic. The processor executes the programmatic control logic. The programmatic control logic includes a prediction application (PA) having control logic that: receives sensor data from the one or more sensors via the input / output ports, and sends control plane data and user plane data to an infrastructure and to the cloud computing servers. The PA control logic further accesses prior knowledge data stored in the memory of the cloud computing servers, performs a Granger causality check on the user plane data and the control plane data, and uses a prediction model to generate a prediction from merged user plane data and control plane data. The control logic of the PA also applies a prediction checker to the prediction from the prediction model and, by means of the PA, allows the host device to make reliable predictions that allow the host device to adapt to conditions of dynamic wireless communication networks even if a best possible model is not fully trained.In yet another aspect of the present invention, the host device further defines a vehicle that communicates with the cloud computing servers via a wireless communication network, and the control logic of the PA further includes control logic for obtaining telematics information, telecommunication information, vehicle telemetry information, and vehicle position information from the one or more sensors.In yet another aspect of the present invention, the method further comprises sending control plane data radio resource control (RRC) messages to an infrastructure including one or more cell towers and to one or more cloud computing servers, and receiving control plane data RRC messages from the one or more cell towers and from the one or more cloud computing servers. The method further comprises sending user plane data to the one or more cell towers and to the one or more cloud computing servers and receiving user plane data from the one or more cell towers and from the one or more cloud computing servers.In yet another aspect of the present invention, the method further comprises accessing prior knowledge data including data obtained from sensors of the host device and from additional devices, data obtained from global positioning system (GPS) satellites, and data obtained from an infrastructure including one or more cell towers, and comparing data control plane data RRC messages from the host device with the prior knowledge data to generate a partial input to the prediction checker.In yet another aspect of the present invention, the method further comprises applying a Granger causality check to the control plane data RRC messages and to the user plane data. The Granger causality check outputs a P value that predicts when a change in the control plane data RRC messages caused a change in the user plane data. The Granger causality check is applied using at least three different P-value based selections including: a threshold P-value selection, a fixed number P-value selection, and a conditional selection. In the threshold P-value selection, all control plane data RRC features having a P-value less than a predetermined threshold are used. In the P-value selection by a fixed number, control plane data RRC features are classified from smallest to largest, and a predetermined number of features having the smallest P-values are selected. In the conditional selection, control plane data RRC features from the P-value selection are compared by threshold and from the P-value selection by a fixed number. When it is determined that the P-value selection has generated a P-value less than the P-value associated with the threshold P-value selection by a predetermined number, the P-value selection indicates a Granger causality between the control plane data RRC features and the user plane data by a predetermined number. When it is determined that the threshold P-value selection has generated a P-value that is less than the P-value associated with the threshold P-value selection by a predetermined number, the threshold P-value selection indicates a Granger causality between the control plane data RRC features and the user plane data. When it is determined that the threshold P-value selection and the fixed number P-value selection values are equal to each other, the threshold P-value selection and the fixed number P-value selection similarly indicate a Granger causality between the control plane data RRC features and the user plane data.In yet another aspect of the present invention, the method further comprises using a Granger causality check output to reduce an amount of data fed into the prediction model from a first amount of data to a second amount of data less than the first and preventing overfeeding of the prediction model, and using a long short-term memory (LSTM) and / or recurrent neural network (RNN) and / or autoregressive integrated sliding average model (ARIMA) model to generate a multivariate prediction from user plane data and control plane data output from the Granger causality check.In yet another aspect of the present invention, the method further comprises online training the PA by: initializing the PA as a null feature model as a reference; and sending control plane messages including Granger causality check result information and a prediction error effective value (RMSE) of a current vehicle-side null feature model to the cloud computing server. Online training of the PA further includes evaluating a vehicle-side null feature model prediction RMSE training speed and predictions, and selectively providing a new model to the host device.In yet another aspect of the present invention, the method further comprises receiving data indicating that a trigger event has occurred, initializing the prediction checker with prior knowledge and with a user plane time series u(n) and a control plane time series C(n) as inputs, and generating a weighted fused statistical prediction according to: p f= w1· p s+ w2· p l, where w1is a first weight applied to a statistical prediction p s generated by a statistical prediction model as applied to the prior knowledge and the user plane time series, w2is a second weight, which is different from the first weight and which is applied to an LSTM prediction p l and p f is a fused prediction which is applied to the fused user plane and control plane data.In yet another aspect of the present invention, the method further comprises applying non-uniform up-sampling in a data collection and time-series formulation for both the user-plane and control-plane data, and reducing bandwidth consumption and processing complexity from a first level to a second level less than the first level by obtaining user-plane and control-plane data only due to the occurrence of a triggering event and at one or more additional times after a predefined time δt after the triggering event. When a value of the control plane time series at time δt after the triggering event is greater than or equal to a value of the control plane time series at a subsequent triggering event, the method causes the PA to increase the control plane time series to collect data due to the occurrence of a subsequent triggering event, and when a value of the control plane time series at time δt after the triggering event is less than the value of the control plane time series at the subsequent triggering event, the method causes the PA to upsample the control plane time series.In yet another aspect of the present invention, a method for predicting a user plane using control plane features and Granger causality includes detecting telematics information, telecommunication information, host device stereometry information, and host device position information with one or more sensors mounted on a host vehicle and communicating with one or more cloud computing servers via a wireless communication network. The method further comprises using one or more controllers, each of the one or more controllers having a processor, memory, and one or more input / output (I / O) ports, the input / output ports in communication with the one or more sensors and the one or more cloud computing servers. The memory stores programmatic control logic. The processor executes the programmatic control logic. The programmatic control logic contains a prediction application (PA). The PA includes control logic for obtaining telematics information, telecommunication information, vehicle telemetry information, and vehicle position information from the one or more sensors. The PA further includes control logic for sending, via the input / output ports, control plane data radio resource control (RRC) messages to an infrastructure including one or more cell towers and to one or more cloud computing servers. The PA further includes control logic for receiving control plane data RRC messages from the one or more cell towers and from the one or more cloud computing servers and sending user plane data to the one or more cell towers and to the one or more cloud computing servers. The PA further includes control logic for receiving user plane data from the one or more cellular towers and from the one or more cloud computing servers. The PA further includes control logic for accessing prior knowledge data stored in the memory of the cloud computing servers, the prior knowledge data including data obtained from sensors of the host device and from additional devices, data obtained from global positioning system (GPS) satellites, and data obtained from an infrastructure including one or more cell towers. The PA further includes control logic for comparing data control plane data RRC messages from the host device with the prior knowledge data to generate a partial input to the prediction checker. The PA further includes control logic for applying a Granger causality check to the control plane data RRC messages and to the user plane data. The Granger causality check outputs a P value that predicts when a change in the control plane data RRC messages caused a change in the user plane data. The Granger causality check is applied using at least three different P-value based selections including: a threshold P-value selection, a fixed number P-value selection, and a conditional selection. In the threshold P-value selection, all control plane data RRC features having a P-value less than a predetermined threshold are used. In the P-value selection by a fixed number, control plane data RRC features are classified from smallest to largest, and a predetermined number of features having the smallest P-values are selected. In the conditional selection, control plane data RRC features from the P-value selection are compared by threshold and from the P-value selection by a fixed number. When it is determined that the P-value selection has generated a P-value less than the P-value associated with the threshold P-value selection by a predetermined number, the P-value selection indicates a Granger causality between the control plane data RRC features and the user plane data by a predetermined number. When it is determined that the threshold P-value selection has generated a P-value that is less than the P-value associated with the threshold P-value selection by a predetermined number, the threshold P-value selection indicates a Granger causality between the control plane data RRC features and the user plane data. When it is determined that the threshold P-value selection and the fixed number P-value selection values are equal to each other, the threshold P-value selection and the fixed number P-value selection similarly indicate a Granger causality between the control plane data RRC features and the user plane data. The PA further includes control logic for using the output of the Granger causality check to reduce an amount of data fed into the prediction model from a first amount of data to a second amount of data that is less than the first and prevents overfeeding of the prediction model. The PA further includes control logic for using a long short-term memory (LSTM) model and / or recurrent neural network (RNN) model and / or an autoregressive integrated sliding average model (ARIMA) model to generate a multivariate prediction from user plane data and control plane data output from the Granger causality check. The PA further includes control logic for using a prediction model to generate a prediction from merged user-level data and control-level data, including: training the PA online by: initializing the PA as a null feature model as a reference; and sending control-level messages including Granger causality check result information and a prediction error effective value (RMSE) of a current vehicle-side null feature model to the cloud computing server. The PA further includes control logic for evaluating a vehicle-side null feature model prediction RMSE training speed and predictions and selectively providing a new model to the host device. The PA further includes control logic for applying a prediction checker to the prediction from the prediction model, including: receiving data indicating that a trigger event has occurred; initializing the prediction checker with prior knowledge and having a user plane time series u(n) and a control plane time series C(n) as inputs; and generating a weighted fused statistical prediction according to: p f= w1·p s+ w2·p l, where w1is a first weight applied to a statistical prediction p s generated by a statistical prediction model as applied to the prior knowledge and the user plane time series, w2is a second weight, which is different from the first weight and which is applied to an LSTM prediction p l and p f is a fused prediction which is applied to the fused user plane and control plane data. The PA further includes control logic for applying non-uniform upsampling in the data collection and time-series formulation for both the user-level and control-level data and reducing bandwidth consumption and processing complexity from a first level to a second level less than the first level by obtaining user-level and control-level data only due to the occurrence of a triggering event and at one or more additional times after a predefined time δt after the triggering event. When a value of the control plane time series at time δt after the triggering event is greater than or equal to a value of the control plane time series at a subsequent triggering event, the method causes the PA to increase the control plane time series to collect data due to the occurrence of a subsequent triggering event, and when a value of the control plane time series at time δt after the triggering event is less than the value of the control plane time series at the subsequent triggering event, the method causes the PA to upsample the control plane time series. The PA further includes control logic for allowing the host device to make reliable predictions that allow the host device to adapt to conditions of dynamic wireless communication networks even if a best possible model is not fully trained.Further areas of applicability will become apparent from the description provided herein. It is to be understood that the specification and specific examples are intended for purposes of illustration only.BRIEF DESCRIPTION OF THE DRAWINGSThe drawings described herein are for illustrative purposes only; and show: FIG. 1 is a schematic diagram of a system for predicting a 5G user plane using control plane features and Granger causality for feature selection according to an aspect of the present invention; FIG. 2 is a flow chart illustrating a logical flow of a Granger causality check portion of the system for predicting a 5G user plane using control plane features and Granger causality for feature selection of FIG. 1, in accordance with an aspect of the present invention; FIG. 3A is a flow chart illustrating a logical flow of a prediction model of the system for predicting a 5G user plane using control plane features and Granger causality for feature selection of FIG. 1, in accordance with an aspect of the present invention; FIG. 3B is a graph depicting exemplary control plane time-series data of the system for predicting a 5G user plane using control plane features and Granger causality for feature selection of FIG. 1, in accordance with an aspect of the present invention; FIG. 4 is a flow chart illustrating a logical flow of a prediction model portion of the system for predicting a 5G user plane using control plane features and Granger causality for feature selection of FIG. 1, in accordance with an aspect of the present invention; FIG. 5 is a schematic diagram illustrating an online training structure of the system for predicting a 5G user level using control level features and Granger causality for feature selection of FIG. 1, according to an aspect of the present invention; FIG. 6 is a flow chart illustrating a logic flow of a predictive checker of the system for predicting a 5G user level using control level features and Granger causality for feature selection of FIG. 1, in accordance with an aspect of the present invention; FIG. 7A is an illustration of a timeline for data collection and time-series formulation in the system for predicting a 5G user plane using control plane features and Granger causality for feature selection of FIG. 1, in accordance with an aspect of the present invention; FIG. 7B is a flow chart illustrating a logic flow of an upsampling process of the system for predicting a 5G user level using control level features and Granger causality for feature selection of FIG. 1, according to an aspect of the present invention; and FIG. 8 is a schematic diagram illustrating the provision of predictions and cloud-based model training in a localized area according to the system for predicting a 5G user plane using control plane features and Granger causality for feature selection of FIG. 1, in accordance with an aspect of the present invention.DETAILED DESCRIPTIONThe following description is merely exemplary in nature.Referring to FIG. 1, a system 10 for predicting a user plane 12 using control plane 14 features and Granger causality for feature selection is shown in graph form. The system 10 includes one or more devices, such as vehicles 16, each having one or more on-board controllers 18. It should be appreciated that the devices may include any of a variety of wirelessly connected host devices such as cellular phones, portable computers such as tablet computers, laptop computers, or the like. However, for convenience, explanation, and refinement, the following description refers to non-limiting examples, where the devices or host devices are vehicles 16.The vehicle-side controllers 18 are non-generalized electronic control devices having a preprogrammed digital computer or processor 20, use a non-transitory computer readable medium or memory 22 to store data such as control logic, software applications, instructions, computer code, data, look-up tables, etc., and include a transmit / receive device or input / output (I / O) ports 24. A computer readable medium includes any type of medium that can be accessed by a computer, such as read only memory (ROM), random access memory (RAM), hard disk drive, compact disk drive (CD), digital video disk drive (DVD), or other type of memory. A "non-transitory" computer readable storage 22 excludes wired, wireless, optical, or other communication links that carry transitory electrical or other signals. A non-transitory computer readable storage 22 includes media where data may be permanently stored and media where data may be stored and later overwritten, such as a rewritable optical disk or an erasable storage device. Computer code includes any type of program code, including source code, object code, and executable code. The processor 20 is configured to execute the code or instructions. In vehicles 12, the controller 18 may be a dedicated Wi-Fi controller or an engine control module, a transmission control module, a body control module, an infotainment control module, etc. The input / output ports 24 are configured to wirelessly communicate using IEEE 802.11x Wi-Fi protocols, cellular protocols such as Global System for Mobile Communication (GSM), Code Division Multiple Access (CDMA), Wireless Local Loop (WLL), General Packet Radio Services (GPRS), 1G, 2G, 3G, 4G Long Term Evolution (LTE), 5G, or the like.The memory 22 may store one or more applications 26. An application 26 is a software program configured to perform a particular function or set of functions. The application 26 may include one or more computer programs, software components, sets of instructions, procedures, functions, objects, classes, instances, related data, or a portion thereof adapted for implementation in suitable computer readable program code. The applications 26 may be stored in the memory 22 of the vehicle-side controllers 18 in the vehicles 16, or in additional or separate memory, such as a storage 22 of a cloud computing device, such as a cloud computing server 28. Examples of applications 26 include audio or video streaming services, games, browsers, social media, and an application for predicting a 5G user plane 12 using features of control plane 14 and Granger causality for feature selection, or the like. Hereinafter, for simplicity and clarity, the application for predicting a 5G user plane 12 using control plane 14 features and Granger causality for feature selection is referred to herein as Statement Application (PA) 30.In several aspects, the vehicles 16 generate, transmit, and receive data while operating. The data may include telematics information, telecommunication information, vehicle telemetry information, position information, data collected from sensors 31 or systems onboard the vehicle 16, or the like. The user plane 12 maintains and communicates the data generated by the vehicles 16, such as data packets processed by protocols such as transmission control protocol (TCP), user diagram protocol (UDP), Internet protocol (IP), and the like. In contrast, in the control plane 14, a bearer protocol such as a radio resource control (RRC) protocol processes signaling messages exchanged between one or more vehicles 16 or between end-user devices such as the one or more vehicles 16 and a base station or infrastructure such as cell towers 32 and the like.Wireless communications, particularly those among GSM, CDMA, WLL, GPRS, 1G, 2G, 3G, 4G and 5G, are subject to a variety of signaling issues that may alter the quality and reliability of the data transmitted and received therein. Accordingly, optimization of user plane 12 metrics such as bandwidth and throughput is important in managing communication network traffic, optimizing communication network performance, and improving user experience.In several aspects, the vehicles 16 send control plane data in the form of RRC messages 34 to an infrastructure such as the cell towers 32 and / or cloud computing servers 28. Similarly, the vehicles 16 receive control plane data RRC messages 34 from cell towers 32, cloud computing servers 28, and the like. When RRC messages 34 of the control plane 14 are received by cloud computing servers 28, cell towers 32, etc., the RRC messages are both added to a prior knowledge database 36 and compared to prior knowledge 36 stored in the memory 22 of the cloud computing servers 28 and / or the cell towers 32 and / or in the vehicle-side memory 22 of the vehicle 16. In several aspects, the prior knowledge 36 is a collection of data obtained from sensors 31 of a host device such as a host vehicle 16, sensors 31 within additional devices or vehicles 16', and sensors 31 located on or acting as part of an infrastructure such as weather satellites, GPS satellites 37, or the like. However, it should be appreciated that the data from additional vehicles 16' in the operating science database 36 is filtered or otherwise constrained based on predefined and / or changeable metrics. Data from additional vehicles 16' can be obtained from these additional vehicles 16' at a variety of locations and over a variety of time periods. It will be appreciated that to make accurate and relevant predictions, the system 10 is best operated using only data from additional vehicles 16' generated by such additional vehicles 16' that are in physical proximity and / or in temporal proximity to the host vehicle 16. Accordingly, data from GPS satellites, weather satellites, or the like may be relevant to the host vehicle 16 for limited time periods or for limited physical distances. Accordingly, the data from GPS and weather satellites are filtered accordingly.The exact definition of "near" may vary depending on the situation, but should be understood to mean that host vehicle 16 and additional vehicles 16' are at a predefined physical and / or temporal distance that allows data from the additional vehicles 16' to remain relevant to host vehicle 16. For example, in a curve on a highway, during seventy miles per hour and variable weather driving, data generated by additional vehicles 16' will have a half-life which is relatively short compared to data from additional vehicles 16' on a straight roadway driving at fifteen miles per hour in sunny, warm, dry weather.Vehicles 16 also send user-level data 38 to an infrastructure such as cellular towers 32 and / or cloud computing servers 28. In several aspects, the user plane data 38 and the RRC messages 34 are subjected to a Granger causality check 40. The Granger causality check is used in supporting the feature selection 42 to generate predictions of user planes 12 for 5G communication. The Granger causality check 40 acts as a predictor, and if Y is "Granger-caused" by X, then it is assumed that past values of X contain information that helps predict the value of Y. The Granger causality check 40 outputs a P value that has a different magnitude or value for different control plane 14 data. In some aspects, if the P value 44 is less than a predefined level of significance, then it is inferred that X has a predictive impact on Y, thereby indicating Granger causality. While the predefined level of significance may vary from application to application, in some examples, the level of significance is less than 0.05.Referring now to FIGS. 2, 3A, and 3B, and with continued reference to FIG. 1, a method 200 of using the Granger causality check 40 to predict features is shown in greater detail in flow chart form. In one example, the Granger causality check 40 is defined as:A feature check is then performed, checking the zero hypothesis that b1=b2=...bm=0. If the null hypothesis is rejected, Granger cause control plane features c(n) will be considered one or more user plane features u(n), where c(n)s will be considered the features used to improve the prediction accuracy of u(n).In some examples, at least three different feature selection methods are used based on the P values of the Granger causality check 40. Specifically, a selection method 202 using threshold value Φ, a selection method 204 using a fixed number, and a conditional selection method 206 are used.The method 202 using threshold Φ selects all c(n)s with a P value 44 less than the threshold Φ 208 according to: where N 1 is the feature number 1 in block 210. It will be appreciated that the threshold Φ is a predefined value that may vary from application to application and from situation to situation, but represents a value chosen to ensure with a predetermined degree of accuracy and reliability that, based on their P values, the c(n)s are accurate predictors.In contrast, method 204 calculates all p values from the data of control plane 14 using a fixed number, and rank 212 the P values from the smallest to the largest. The method 204 using a fixed number then selects a predetermined number of the features having the smallest P values. In one example, the method 204 selects, using a fixed number, the three (3) features having the smallest P values, however, the exact set of features that are selected may vary. The selection process may be represented as a selection of an N2number of c(n)swith the lowest P values according to: where N2is the feature number 2 in block 214.Finally, the conditional method 206 compares the values of N1 and N2 in block 216 to determine whether a Granger causality is present. More specifically:Accordingly, C_ 3 is theoretically a Granger cause of u(n) among the limits of the available feature numbers. That is, if N 1<N 2, then the method 204 has provided the most accurate result using a fixed number, while if N 2<N 1, then the method 202 has provided the most accurate result using threshold value Φ. Similarly, if N 2=N 1, then both the threshold Φ selection method 202 and the fixed number selection method 204 have the same probability of generating accurate results.Referring now in particular to FIGS. 3A and 3B and with continued reference to FIGS. 1 and 2, twenty (20) different time-series data points of the control plane 14 are collected and represented as the function c(t). Granger causality checks [become] 40 determine the probabilities of the c(n)scaused by granger in a user-level time series u(n). As previously stated, the lower a P value 44, the higher the probability that u(n) is Granger-caused by c(n) Granger-, the higher the probability that Granger can therefore be represented as the function P(GC|p) ≅ f(p) in block 46, where f is a decreasing function.Referring to FIG. 4, and with continued reference to FIGS. 1-3, the system 10 uses a prediction model 300 to predict short-term data rates 301. Prediction model 300 starts at block 302 using user plane 12 data [represented as u(i:i+1)] and control plane 14 data [represented as C(i:i+1)] as inputs. In several aspects, the Granger causality check 40 is used to check the input data prior to use in the prediction model 300 to reduce the potential of overfeeding the prediction model 300. Overfeeding the prediction model 300 may cause the prediction model 300 to generate an output that is effectively meaningless because the prediction model 300 itself does not determine which input data is actually useful for the vehicle 16 or the wireless communication network functionality. That is, the Granger causality check 40 reduces an amount of data input to the prediction model 300 from a first amount of data to a second amount of data that is less than the first amount of data to avoid overfeeding the prediction model 300. In addition to preventing overfeeding-a factor that may degrade look-ahead performance of the system 10-the Granger causality check 40 provides the capacity to choose "relevant" input variables, thereby improving the look-ahead accuracy of the model as compared to models without such inputs.The data of the user and control plane 12, 14 are normalized 304 and processed by a prediction algorithm 306. The prediction algorithm 306 shown is a long short-term memory (LSTM), however, the LSTM may be replaced with other feature-based time-series prediction models such as a recurrent neural network (RNN), an autoregressive integrated moving average (ARIMA), or the like. Prediction model 300 then performs inverse normalization 308 and generates an output 310 of user plane 14 [represented as u'(i+1)], which is then compared to a target output 312 of user plane 14 [represented as u(i+1)] to generate an effective error (RMSE) 314. In several aspects, the RMSE 314 defines a metric used to measure a prediction error between the prediction output 310 of the user plane 14 and the target output 312 which is ground truth.Referring to FIG. 5, and with continued reference to FIGS. 1-4, an online training structure 400 for the PA 30 is schematically shown. The PA 30 is initialized as a null feature model (0-feature model) 402 operating at the controller 18 of the host vehicle 16. The null feature model 402 serves as a reference against which further features are subsequently compared. The host vehicle 16 sends current control plane 14 messages containing Granger causality check 40 and current vehicle null feature model 402 prediction RMSE 314 to the cloud computing server 28. Depending on comparisons of the prediction RMSE 314 and training speed evaluations, the cloud computing server 28 decides whether and when to provide a new model to the vehicle 16 and thereby replace 404 the stream model with a new model. Training based on cloud computing server 28 allows host vehicle 16 to make reliable predictions even if the best possible model is not fully trained. It will be appreciated that the set of features in a given model correlates with relative speeds of training and provisioning processes. That is, a single feature model 406 may be trained at a faster rate than a two feature model (2-feature model) 408, etc. In other example applications of the online training structure 400, results of the single feature model 406 are compared to the two feature model 408 through an N feature model 410. If the two-feature model 408 achieves an empirically better result than the single-feature model 406, then the two-feature model 408 is provided to the host vehicle 16. Similarly, if the N feature model 410 achieves a better result than an N-1 feature model (not specifically shown), then the N feature model 410 is provided to the host vehicle 16. It will be appreciated that training an N feature model 410 is computationally more complex and therefore slower than the single feature model 406, the two feature model 408, etc. However, the N feature model 410 is more accurate and reliable than the N-1 feature model because the N feature model 410 is optimized from Granger causality checks 40 of each of the sub-N feature models 410. Accordingly, the N feature model 410 is effectively an optimized fine-tuned feature model for the host vehicle 16.Referring now to FIG. 6, and with continued reference to FIGS. 1-5, a predictive checker 500 is shown in flow chart form. Predictive checker 500 effectively employs insights obtained from field experimentation to detect and mitigate the effect of predictive outliers on the predictions of PA 30 as time progresses. In several aspects, the prediction verifier 500 is initialized upon receipt of data indicative of the occurrence of a trigger event (n) and generates a prediction (m). Triggering events (n) may be any of a variety of different types of changes to user and / or control plane 12, 14 situations. In a non-limiting example, changes to the control plane 14 may directly change or affect the state of the user plane 12 when a handover event occurs between multiple cell towers 32. That is, during handover between cell towers 32, a bandwidth of communications between the host vehicle 16 and the cell towers 32 may decrease from a first level to a second level less than the first. Accordingly, if line-of-sight communication between the host vehicle 16 and cellular towers 32 or the like may be complicated by the presence of large buildings, tunnels, mountains, etc., thereby causing a change in the activity of the control plane 14. That is, an RSRP (reference signal reception power), an RSRQ (reference signal reception quality), and / or an RSSI (received signal strength indicator) may be negatively affected by a blocked line of sight, handover of cellular towers 32, and the like. In another example, when a speed of the host vehicle 16 is low, multiple or frequent handoffs between cell towers 32 may occur at a small physical distance, thereby causing bandwidth droop and changes in user plane 12 experience. In yet another example, when a high bandwidth demand application, e.g., a video streaming service application accessed via a human machine interface (HMI) of a host vehicle 16 or via user equipment in electronic communication with the controller 18, an amount of bandwidth released for use by other systems may drastically change in the control plane 14, resulting in changes in functionality of the user plane 12. Accordingly, the prediction checker 500 checks and adjusts the prediction results of the LSTM prediction algorithm 306.Accordingly, the prediction checker 500 uses prior knowledge 36 as well as the user plane time series u(n) and the control plane time series C(n) as inputs. In particular, user-level time series u(n) and prior knowledge 36 are used as inputs to a statistical prediction model 502. The statistical prediction model 502 may be any of a variety of different statistical prediction model types including, but not limited to, zero hypothesis (NHST) statistical testing models, multivariate Kalman filters, multivariate Bayesian analyses, or the like. The statistical prediction model 502 generates a statistical prediction 504 or statistical prediction result p s. Accordingly, the control plane time series C(n) and user plane time series u(n) data are used as inputs to the LSTM prediction algorithm 306. The LSTM prediction algorithm 306 generates an LSTM statistical prediction 506 or an LSTM prediction result p l. The statistical prediction results p s are used to reduce further errors of the LSTM prediction results p l and to generate a final prediction p f508 defined by: where prediction weights (w1 and w2) 510 are adjusted to minimize the following error effective value loss (MSE) function:The weights w1and w2are then updated using a gradient descent algorithm according to the following:Then, the gradients themselves are calculated according to the following: where T is the ground truth and a is a learning rate. The checking process is carried out at specific trigger events (n), wherein prior knowledge 36 has a substantial influence on the change in the data rate. It will be appreciated that the weights w1, w2, w(j-1), w(j) may vary from application to application and from model to model, but it should be understood that they support the prediction checker 500 and the PA 30 by selecting which of the single feature, two feature or up to N feature model 406, 408, 410 is the appropriate model to address a given trigger event(s).Referring to FIGS. 7A and 7B, to further improve the accuracy of predictions, the PA 30 uses non-uniform or non-uniform upsampling in a data collection and time-series formulation.Specifically, the control plane 14 data is collected from RRC messages at the client or host vehicle 16. Control plane 14 key information (C) is extracted from the RRC messages and monitored according to: while the user plane time series can be represented as: where Thput(n̅:n) is the system throughput or bandwidth. In some aspects, when c(n)=1, C is detected, the C that has been detected may be handover between cellular towers 32 or the like. While uniform scanning to some extent may be used, it has several disadvantages in the present context. Specifically, with uniform sampling, it is difficult to align sample points with the exact occurrence of control plane 14 events, thereby introducing additional measurement uncertainty and increasing the possibility of errors. Likewise, a high sampling frequency with uniform sampling results in a large volume of data, high bandwidth consumption and increased computational complexity in the prediction model. Non-uniform or non-uniform sampling offers certain advantages, namely the ability to sample based on the occurrence of control plane 14 events. However, because non-uniform or non-uniform scanning is triggered only due to the occurrence of a particular control plane 14 event, changes to the user plane 12 may be lost between the occurrence of multiple control plane 14 trigger events. That is, U(n) represents an average data rate between two successive events of the control plane 14, but fails to accommodate changes that occur therebetween. Accordingly, the PA 30 applies up-scanning at locations of substantial change of the user plane 12.In a data collection and time-series formulation, the PA 30 samples the user-plane time-series u(n) twice using an additional sampling point at a time δt after each of the events of the control plane 14. δt is obtained by estimating the time at which the most significant fluctuation of throughput occurs following a particular event of the control plane 14. Substituting the δt-based sampling data into the sampling procedure improves prediction accuracy. Referring now in particular to FIG. 7B, an upsampling method 600 is shown in more detail. In block 602, the PA 30 receives data of a control plane time series c(n). In block 604, the PA 30 appends the upsampled control plane time series c_upsampled to the control plane time series c(n). In block 606, the PA 30 compares the value of the control plane time series c(n) at time δt after a triggering event (n) to the value of the control plane time series c(n) at a subsequent triggering event (n+1). When the value of the control plane time series c(n) at time δt after the triggering event (n) is greater than or equal to the value of the control plane time series c(n) at the subsequent triggering event (n+1), the control plane time series c(n) is increased to the collected data after the occurrence of a subsequent triggering event n++. However, if the value of the control plane time series c(n) at time δt after the trigger event (n) is less than the value of the control plane time series c(n) at the subsequent trigger event (n+1), an up-sampled control plane time series c_upsampled is obtained in block 608 and then applied to the control plane time series c(n).Referring to FIG. 8, and with continued reference to FIGS. 1-7B, models generated by the PA 30 may be shared between a host vehicle 16 and additional vehicles 16' having similar characteristics. For example, additional vehicles 16', traveling along the same or substantially similar paths at a particular physical location and time of day, may have the same prediction model provided to each of the similarly located additional vehicles 16', and vice versa. That is, one or more of multiple vehicles 16 may generate a prediction of the user plane 12 and / or the control plane 14 by the PA 30, which is then uploaded to the cloud computing server 28. Cloud computing server 28 may then provide the prediction of user plane 12 and / or control plane 14 to additional vehicles 16' in the same environment. For each vehicle 16 that uploads data to the cloud computing server 28, additional data is used to further train the model in the cloud. That is, while each vehicle 16 is reporting to the cloud computing server 28, the upload data rate, the GPS data, the speed, and the like are uploaded and used as additional training information to further refine and increase the accuracy of predictions generated by the PA 30.A system and method for predicting a user plane 12 using control plane 14 features and Granger causality for feature selection of the present invention provides several advantages. These include the ability to accurately accommodate complex causal relationships between control plane 14 information and user plane 12 metrics in a highly dynamic 5G environment in which changes to control plane 14 can directly affect user plane 12. The Granger causality provides a robust and reasonable means for choosing features for predictive models in a practical, reliable and computationally less impacted manner, while optimizing performance, improving cellular communication reliability and user experience, functioning well in dynamic environments, and being easily adaptable to new and existing platforms without increasing system complexity.

Claims

A system (10) for predicting a user plane (12) using control plane features and Granger causality, comprising: a host device (16) having one or more sensors (31), wherein the one or more sensors (31) detect telematics information, telecommunication information, host device stereometry information, and host device position information; one or more cloud computing servers (28); one or more controllers (18), each of the one or more controllers (18) having a processor (20), a memory (22), and one or more input / output (I / O) ports (24), the input / output ports (24) being in communication with the one or more sensors (31) and the one or more cloud computing servers (28); the memory (22) storing programmatic control logic; the processor (20) executing the programmatic control logic; the programmatic control logic includes a prediction application (PA), and the PA comprises: first control logic for obtaining sensor data from the one or more sensors (31); second control logic for sending control plane data (34) and user plane data (38) to an infrastructure and to the cloud computing servers (28) via the input / output ports (24); third control logic for accessing prior knowledge data stored in the memory (22) of the cloud computing servers (28); fourth control logic for performing a Granger causality check (40) on the user plane data (38) and the control plane data (34); fifth control logic for using a prediction model (300) to generate a prediction from merged user plane data (38) and control plane data (34); and sixth control logic for applying a prediction checker (500) to the prediction of the prediction model (300); wherein the PA allows the host device (16) to make reliable predictions that allow the host device (16) to adapt to conditions of dynamic wireless communication networks even if a best possible model is not fully trained.The system (10) of claim 1, wherein the host device (16) further comprises a vehicle (16) that communicates with the cloud computing servers (28) via a wireless communication network; and the first control logic further comprises: obtaining telematics information, telecommunication information, vehicle telemetry information, and vehicle position information from the one or more sensors (31).The system (10) of claim 1, wherein the second control logic further comprises: sending control plane data radio resource control (RRC) messages to an infrastructure including one or more cell towers (32) and to one or more cloud computing servers (28); receiving control plane data RRC messages (34) from the one or more cell towers (32) and from the one or more cloud computing servers (28); sending user plane data (38) to the one or more cell towers (32) and to the one or more cloud computing servers (28) and receiving user plane data (38) from the one or more cell towers (32) and from the one or more cloud computing servers (28).The system (10) of claim 1, wherein the third control logic further comprises: accessing prior knowledge data including data obtained from sensors (31) of the host device (16) and from additional devices, data obtained from global positioning system (GPS) satellites (37), and data obtained from an infrastructure including one or more cellular towers (32); and comparing data control plane data RRC messages (34) from the host device (16) with the prior knowledge data to generate a partial input to the prediction checker (500).The system (10) of claim 3, wherein the fourth control logic further comprises: applying a Granger causality check (40) to the control plane data RRC messages (34) and to the user plane data (38), wherein the Granger causality check (40) outputs a P value that predicts when a change in the control plane data RRC messages (34) caused a change in the user plane data (38); wherein the Granger causality check (40) is applied by means of at least three different selections based on the P value, including: a P value selection by means of a threshold value, a P value selection by means of a fixed number and a conditional selection, wherein in the P value selection by means of threshold value all control plane data RRC features are used which have a P value which is smaller than a predetermined threshold value; in the P value selection by means of a fixed number of control plane data RRC features are classified from the smallest to the largest and a predetermined number of features having the smallest P values is selected; and in the conditional selection, control plane data RRC features from the thresholded P-value selection and from the thresholded P-value selection are compared by a fixed number; wherein if it is determined that the thresholded P-value selection has generated a P-value that is less than the thresholded P-value selection, the thresholded P-value selection indicates a Granger causality between the control plane data RRC features and the user plane data (38); if it is determined that the thresholded P-value selection has generated a P-value that is less than the P-value, which is assigned to the P-value selection by a fixed number, the P-value selection by threshold indicates a Granger causality between the control plane data RRC features and the user plane data (38), and when it is determined that the P-value selection by threshold and the P-value selection values of a fixed number are equal to each other, the P-value selection by threshold and the P-value selection by a fixed number similarly indicate a Granger causality between the control plane data RRC features and the user plane data (38).The system (10) of claim 5, wherein the fifth control logic further comprises: using an output of the Granger causality check (40) to decrease an amount of data fed into the prediction model (300) from a first amount of data to a second amount of data less than the first and preventing overfeeding of the prediction model (300); using a long short-term memory (LSTM) model and / or recurrent neural network (RNN) model and / or an autoregressive integrated sliding average model (ARIMA) to generate a multivariate prediction from user plane data (38) and control plane data (34) output from the Granger causality check (40).The system (10) of claim 6, further comprising: control logic for online training the PA by: initializing the PA as a null feature model (402) as a reference; sending control plane messages including Granger causality check result information and a prediction error effective value (RMSE) of a current vehicle-side null feature model (402) to the cloud computing server (28); evaluating a vehicle-side null feature model prediction RMSE training speed and predictions and optionally providing a new model to the host device (16).The system (10) of claim 6, wherein the sixth control logic further comprises: receiving data indicating that a trigger event has occurred; initializing the prediction checker (500) with prior knowledge and having a user plane time series u(n) and a control plane time series C(n) as inputs; and generating a weighted fused statistical prediction according to: p f= w1 · p s+ w2 · p l, where w1 is a first weight applied to a statistical prediction p s generated by a statistical prediction model (502) as applied to the prior knowledge and the user plane time series, w2 is a second weight, which is different from the first weight and which is applied to an LSTM prediction p l and p f is a fused prediction which is applied to the fused user plane and control plane data (38,34).The system (10) of claim 1, wherein the control logic of the PA further includes: control logic that applies non-uniform upsampling in a data collection and time series formulation for both the user-level and control-level data (38, 34).The system (10) of claim 9, wherein the non-uniform up-sampling further comprises: control logic that reduces bandwidth consumption and processing complexity from a first level to a second level less than the first level by obtaining user-level and control-level data (38, 34) only due to the occurrence of a triggering event and at one or more additional times after a predefined time δt after the triggering event; When a value of the control plane time series at time δt after the triggering event is greater than or equal to a value of the control plane time series at a subsequent triggering event, then cause the PA to increase the control plane time series to collect data due to the occurrence of a subsequent triggering event, and when a value of the control plane time series at time δt after the triggering event is less than the value of the control plane time series at the subsequent triggering event, then cause the PA to upsample the control plane time series.

Images