METHOD FOR ESTIMATE THE SIGNAL-TO-DISTURBANCE AND NOISE RATIO

Abstract

A method for estimating the signal-to-noise ratio (SINR) in a wireless network environment is provided. The method may include collecting multiple datasets of wireless connections across the wireless network environment. Each of the multiple datasets contains at least one measured environmental parameter and a measured SINR. The method may further include generating a regression model based at least partially on the multiple datasets of wireless connections. The regression model is configured to determine an estimated SINR based on the at least one measured environmental parameter. The method may also include training a machine learning algorithm for SINR optimization to determine an optimized estimated SINR based at least partially on the estimated SINR.

Description

INTRODUCTION

[0001] The present invention relates to wireless network communication and, in particular, to a method according to the preamble of claim 1 for estimating the signal-to-interference and noise ratio (SINR) in a wireless network environment, as is essentially known from WO 2019 / 211 792 A1.

[0002] Further prior art can also be found in WO 2022 / 069 061 A1.

[0003] To increase occupant awareness and comfort, vehicles may be equipped with various sensors and systems, such as awareness sensors (e.g., camera, radar, ultrasonic distance sensors, and / or the like), microphones, navigation systems, advanced driver assistance systems (ADAS), automated driving systems (ADS), and / or the like. Such sensors and systems may generate data that is transmitted wirelessly using a cellular network infrastructure. Therefore, an increasing number of vehicles may utilize the cellular network infrastructure to transmit critical data regarding occupant awareness and comfort. A deviation in environmental conditions, such as the presence of tall buildings and fluctuations in a number of connected devices (e.g.,connected vehicles), but can lead to a change in the signal-to-interference and noise ratios even over relatively small geographical areas.

[0004] Consequently, while current systems and methods for estimating the signal-to-interference and noise ratio (SINR) in a wireless network environment serve their intended purpose, there is a need for a new and improved system and method for estimating the signal-to-interference and noise ratio (SINR) in a wireless network environment with a high geographical resolution. SUMMARY

[0005] According to the invention, a method for estimating the signal-to-interference and noise ratio (SINR) in a wireless network environment is provided, which is characterized by the features of claim 1. The method comprises collecting a plurality of wireless connection data sets about the wireless network environment. Each of the plurality of wireless connection data sets includes at least one measured environmental parameter and a measured SINR. The method may further include generating a regression model based at least in part on the plurality of wireless connection data sets. The regression model is configured to determine an estimated SINR based on the at least one measured environmental parameter.The method may further include training a SINR optimization machine learning algorithm to determine an optimized estimated SINR based at least in part on the estimated SINR. Collecting the plurality of wireless connection data sets further includes receiving the plurality of wireless connection data sets from one or more wireless devices. The at least one measured environmental parameter of each of the plurality of wireless connection data sets includes at least one of a channel quality indication (CQI) value, a weather condition value, and an environmental complexity value. The environmental complexity value quantifies a number of obstacles and / or an average height of the obstacles that impede wireless transmissions.

[0006] According to another aspect of the present invention, the environmental complexity value is determined by each of the one or more wireless devices using a machine learning model to identify environmental complexity.

[0007] According to another aspect of the present invention, the one or more wireless devices include at least one vehicle equipped with a vehicle communication system.

[0008] According to another aspect of the present invention, training the SINR optimization machine learning algorithm may further include determining a plurality of tested SINR deviations between the estimated SINR determined by the regression model and the measured SINR in each of the plurality of wireless link data sets. Training the SINR optimization machine learning algorithm may further include training the SINR optimization machine learning algorithm based at least in part on the plurality of tested SINR deviations. The SINR optimization machine learning algorithm is trained to receive the at least one measured environmental parameter and the estimated SINR as an input and to provide the optimized estimated SINR as an output.

[0009] According to another aspect of the present invention, training the SINR optimization machine learning algorithm may further include training the SINR optimization machine learning algorithm to account for edge cases using multiple simulated wireless network environments.

[0010] According to another aspect of the present invention, training the SINR optimization machine learning algorithm to account for edge cases using multiple simulated wireless network environments may further include generating multiple simulated wireless network environments using a computer simulation. Training the SINR optimization machine learning algorithm to account for edge cases using multiple simulated wireless network environments may further include training the SINR optimization machine learning algorithm based at least in part on the multiple simulated wireless network environments.

[0011] According to another aspect of the present invention, generating the plurality of simulated wireless network environments may further include initializing the plurality of simulated wireless network environments. Each of the plurality of simulated wireless network environments is defined by at least one simulated environment parameter. Generating the plurality of simulated wireless network environments may further include adding a plurality of simulated wireless nodes to each of the plurality of simulated wireless network environments to form a plurality of simulated wireless network environments. Generating the plurality of simulated wireless network environments may further include measuring one or more simulated SINRs within each of the plurality of simulated wireless network environments.

[0012] According to another aspect of the present invention, training the SINR optimization machine learning algorithm based at least in part on the plurality of simulated wireless network environments may further include determining one or more estimated SINRs for each of the plurality of simulated wireless network environments using the regression model based at least in part on the at least one simulated environment parameter of each of the plurality of simulated wireless network environments. Training the SINR optimization machine learning algorithm based at least in part on the plurality of simulated wireless network environments may further include comparing the one or more estimated SINRs to the one or more simulated SINRs for each of the plurality of simulated wireless network environments to determine a plurality of simulated SINR deviations.Training the SINR optimization machine learning algorithm based at least in part on the plurality of simulated wireless network environments may further include training the SINR optimization machine learning algorithm based at least in part on the at least one simulated environment parameter of each of the plurality of simulated wireless network environments and the plurality of simulated SINR deviations.

[0013] Further described is a system for estimating the signal-to-interference and noise ratio (SINR) in a wireless network environment. The system may include a plurality of wireless nodes. The system may further include one or more central computers in electrical communication with the plurality of wireless nodes. The one or more central computers are programmed to collect a plurality of wireless link data sets about the wireless network environment using the plurality of wireless nodes. Each of the plurality of wireless link data sets includes at least one measured environmental parameter and a measured SINR. The one or more central computers are further programmed to generate a regression model based at least in part on the plurality of wireless link data sets.The regression model is configured to determine an estimated SINR based on the at least one measured environmental parameter. The one or more central computers are further programmed to train a machine learning algorithm for SINR optimization to determine an optimized estimated SINR based at least in part on the estimated SINR.

[0014] According to another aspect of the present invention, the one or more central computers are further programmed to receive the plurality of wireless connection data sets from the plurality of wireless nodes to collect the plurality of wireless connection data sets. The plurality of wireless connection data sets are collected by one or more wireless devices wirelessly communicating with the plurality of wireless nodes. The one or more wireless devices include at least one vehicle equipped with a vehicular communication system. The at least one measured environmental parameter of each of the plurality of wireless connection data sets includes at least one of: a channel quality indication (CQI) value, a weather condition value, and an environmental complexity value.

[0015] According to another aspect of the present invention, the one or more central computers are programmed to determine a plurality of tested SINR deviations between the estimated SINR determined by the regression model and the measured SINR in each of the plurality of wireless link data sets to train the SINR optimization machine learning algorithm. To train the SINR optimization machine learning algorithm, the one or more central computers are further programmed to train the SINR optimization machine learning algorithm based at least in part on the plurality of tested SINR deviations. The SINR optimization machine learning algorithm is trained to receive the at least one measured environmental parameter and the estimated SINR as an input and to provide the optimized estimated SINR as an output.

[0016] According to another aspect of the present invention, the one or more central computers are further programmed to train the SINR optimization machine learning algorithm to consider edge cases using a plurality of simulated wireless network environments to train the SINR optimization machine learning algorithm.

[0017] According to another aspect of the present invention, the one or more central computers are further programmed to generate a plurality of simulated wireless network environments using a computer simulation to train the SINR optimization machine learning algorithm to account for edge cases using the plurality of simulated wireless network environments. To train the SINR optimization machine learning algorithm to account for edge cases using the plurality of simulated wireless network environments, the one or more central computers are further programmed to train the SINR optimization machine learning algorithm based at least in part on the plurality of simulated wireless network environments.

[0018] According to another aspect of the present invention, the one or more central computers are further programmed to initialize the plurality of simulated wireless network environments to generate the plurality of simulated wireless network environments. Each of the plurality of simulated wireless network environments is defined by at least one simulated environment parameter. The at least one simulated environment parameter for each of the plurality of simulated wireless network environments is determined by sampling a plurality of probability distributions corresponding to each of the at least one simulated environment parameter. To generate the plurality of simulated wireless network environments, the one or more central computers are further programmed to add a plurality of simulated wireless nodes to each of the plurality of simulated wireless network environments to form a plurality of simulated wireless network environments.To generate the plurality of simulated wireless network environments, the one or more central computers are further programmed to measure one or more simulated SINRs within each of the plurality of simulated wireless network environments.

[0019] According to another aspect of the present invention, the one or more central computers are further programmed to determine one or more estimated SINRs for each of the plurality of simulated wireless network environments using the regression model based at least in part on the at least one simulated environment parameter of each of the plurality of simulated wireless network environments to train the SINR optimization machine learning algorithm based at least in part on the plurality of simulated wireless network environments.To train the SINR optimization machine learning algorithm based at least in part on the plurality of simulated wireless network environments, the one or more central computers are further programmed to compare the one or more estimated SINRs to the one or more simulated SINRs for each of the plurality of simulated wireless network environments to determine a plurality of simulated SINR deviations. To train the SINR optimization machine learning algorithm based at least in part on the plurality of simulated wireless network environments, the one or more central computers are further programmed to train the SINR optimization machine learning algorithm based at least in part on the at least one simulated environment parameter of each of the plurality of simulated wireless network environments and the plurality of simulated SINR deviations.

[0020] Further described is a method for estimating the signal-to-interference and noise ratio (SINR) in a wireless network environment. The method may include receiving a plurality of wireless connection data sets from one or more wireless devices. Each of the plurality of wireless connection data sets includes at least one measured environmental parameter and a measured SINR. The at least one measured environmental parameter of each of the plurality of wireless connection data sets includes at least one of a channel quality indication (CQI) value, a weather condition value, and an environmental complexity value. The method may further include generating a regression model based at least in part on the plurality of wireless connection data sets. The regression model is configured to determine an estimated SINR based on the at least one measured environmental parameter.The method may further include determining a plurality of tested SINR deviations between the estimated SINR determined by the regression model and the measured SINR in each of the plurality of wireless link data sets. The method may further include training a SINR optimization machine learning algorithm based at least in part on the plurality of tested SINR deviations. The SINR optimization machine learning algorithm is trained to receive the at least one measured environmental parameter and the estimated SINR as an input and to provide an optimized estimated SINR as an output.

[0021] According to another aspect of the present invention, training the SINR optimization machine learning algorithm may further include generating a plurality of simulated wireless network environments using a computer simulation. Training the SINR optimization machine learning algorithm may further include training the SINR optimization machine learning algorithm to account for edge cases based at least in part on the plurality of simulated wireless network environments.

[0022] According to another aspect of the present invention, training the SINR optimization machine learning algorithm based at least in part on the plurality of simulated wireless network environments may further include determining one or more estimated SINRs for each of the plurality of simulated wireless network environments using the regression model based at least in part on at least one parameter of the simulated environment of each of the plurality of simulated wireless network environments. Training the SINR optimization machine learning algorithm based at least in part on the plurality of simulated wireless network environments may further include comparing the one or more estimated SINRs to one or more simulated SINRs for each of the plurality of simulated wireless network environments to determine a plurality of simulated SINR deviations.Training the SINR optimization machine learning algorithm based at least in part on the plurality of simulated wireless network environments may further include training the SINR optimization machine learning algorithm based at least in part on the at least one simulated environment parameter of each of the plurality of simulated wireless network environments and the plurality of simulated SINR deviations.

[0023] Further areas of application will become apparent from the description provided herein. It should be understood that the description and specific examples are provided for illustrative purposes only. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The drawings described here are for illustrative purposes only; they show: Fig. 1 is a schematic diagram of a system for estimating the signal-to-interference and noise ratio (SINR) in a wireless network environment according to an exemplary embodiment; Fig. 2 is a flowchart of a method for estimating the signal-to-interference and noise ratio (SINR) in a wireless network environment according to an exemplary embodiment; Fig. 3 is a schematic diagram of a simulated wireless network for a computer simulation according to an exemplary embodiment; and Fig. 4 is a flowchart of a method for using a regression model and a machine learning algorithm for SINR optimization according to an exemplary embodiment. DETAILED DESCRIPTION

[0025] According to aspects of the present invention, variations in environmental conditions, including, for example, the presence of tall buildings and fluctuations in a number of connected devices (e.g., connected vehicles), can result in changes in signal-to-interference and noise ratios even over relatively small geographic areas. Accordingly, the present invention provides a new and improved system and method for SINR estimation that enables estimation of SINR at a high-resolution local level.

[0026] The following description is for illustrative purposes only.

[0027] In Fig. 1 illustrates a system for estimating the signal-to-interference and noise ratio (SINR) in a wireless network environment, indicated generally by reference numeral 10. System 10 generally includes one or more central computers 12 and a plurality of wireless nodes 14.

[0028] The one or more central computers 12 are used to implement a method 100 for estimating the signal-to-interference and noise ratio (SINR) in a wireless network environment, as described below. The one or more central computers 12 include at least one controller 16, a central computer communication system 18, and a database 20.

[0029] The controller 16 includes at least one processor 22 and a non-transitory computer-readable storage device or non-transitory computer-readable storage media 24. The processor 50 may be a custom-built or off-the-shelf processor, a central processing unit (CPU), a graphics processing unit (GPU), an auxiliary processor among a plurality of processors associated with the one or more central computers 12, a semiconductor-based microprocessor (in the form of a microchip or chipset), a macroprocessor, a combination thereof, or generally an apparatus for executing instructions.

[0030] The computer-readable storage device or storage media 24 may include, for example, volatile and non-volatile memory in read-only memory (ROM), random access memory (RAM), and persistent memory (RAM). A RAM is persistent or non-volatile memory that can be used to store various operating variables while the processor 22 is powered off. The computer-readable storage device or storage media 24 may be implemented using a number of storage devices, such as, for example,PROMs (programmable read-only memory), EPROMs (electrical PROMs), EEPROMs (electrically erasable PROMs), flash memory, or other electrical, magnetic, optical, or combination storage devices capable of storing data, some of which represent executable instructions used by the one or more central computers 12 to carry out the method 100.

[0031] The central computer communication system 18 is used by the controller 16 to communicate with other systems external to the one or more central computers 12. The central computer communication system 18 includes, for example, vehicle-to-vehicle (“V2I”) communication capabilities. According to certain embodiments, the central computer communication system 18 is a wireless communication system configured to communicate over a wireless local area network (WLAN) using the IEEE 802.11 standards or using cellular data communication (e.g., using the GSMA standards such as SGP.02, SGP.22, SGP.32, and the like). Accordingly, the central computer communication system 18 may further include an embedded universal integrated circuit card (eUICC) configured to implement at least one cellular connectivity capability configuration profile, e.g.,an embedded subscriber identity module (eSIM) profile. The central computer communication system 18 is further configured to communicate via a personal area network (e.g., BLUETOOTH), near field communication (NFC), and / or any additional type of radio frequency communication.

[0032] Accordingly, the central computer communication system 18 may include one or more antennas and / or one or more communication transceivers for receiving and / or transmitting signals. The central computer communication system 18 is configured to wirelessly transmit information between the one or more central computers 12 and external communication networks, such as the Internet. The central computer communication system 18 is in electrical communication with the controller 16. It should be appreciated that the central computer communication system 18 may be integrated with the one or more central computers 12 (e.g., may be located on the same circuit board with the one or more central computers 12 or may otherwise be a part of the one or more central computers 12).

[0033] The database 20 is used to store information about the SINR in the wireless network environment, as discussed in more detail below. According to an exemplary embodiment, the database 20 includes one or more mass storage devices, such as hard disk drives, magnetic tape drives, magneto-optical disk drives, optical disks, solid-state drives, and / or additional devices operable to store data persistently and in a machine-readable manner. According to some examples, the one or more mass storage devices may be configured to provide redundancy in the event of a hardware failure and / or data corruption, for example, using a redundant array of independent disks (RAID). According to one non-limiting example, the controller 16 may implement software, such asa database management system (DBMS) that enables the data stored in the one or more mass storage devices to be organized and accessed. The database 20 is in electrical communication with the controller 16.

[0034] The controller 16 is in electrical communication with the central computer communication system 18 and the database 20. According to an exemplary embodiment, the electrical communication is established using, for example, a CAN network, a FLEXRAY network, a local area network (e.g., Wi-Fi, Ethernet, and the like), a serial peripheral interface (SPI) network, or the like. It should be recognized that various additional wired and wireless technologies and communication protocols may be used to communicate with the controller 16.

[0035] The plurality of wireless nodes 14 are wireless telecommunications nodes. A wireless node is a location that includes electronic communication devices (e.g., antennas, transceivers, digital signal processors, control electronics, and / or the like) configured to connect a user equipment (UE) to a wireless network (e.g., a cellular network). A UE includes, for example, wireless devices, mobile devices (e.g., cellular phones, smartphones, laptops, and / or the like), Internet of Things (IoT) devices, connected vehicles, and / or the like.

[0036] According to an exemplary embodiment, each of the plurality of wireless nodes 14 includes one or more antennas and a base transceiver station (BTS). The one or more antennas are used to transmit signals to and receive signals from the UE. The one or more antennas may include various types of antennas, e.g., dipole antennas such as dipole turnstile antennas, dipole corner reflector antennas, and dipole microstrip antennas. The one or more antennas may also include monopole antennas such as monopole whip antennas. It should be appreciated that many additional types of antennas, such as array antennas, loop antennas, and aperture antennas, may be used to transmit radio signals to and receive signals from the UE.

[0037] The base transceiver station (BTS) is used to receive and transmit radio frequency (RF) signals using the one or more antennas. According to one non-limiting example, the BTS includes one or more RF transceivers, wherein each of the one or more RF transceivers supports multiple simultaneous connections to the UE. According to another non-limiting example, the BTS may further include additional signal processing devices, such as devices for encrypting and / or decrypting communications, spectral filters, multiplexers, amplifiers, and / or the like. According to one exemplary embodiment, the BTS is in electrical communication with a core network infrastructure. According to one non-limiting example, the connection to the core network infrastructure may be established using, for example, fiber optics, microwave links, satellite links, and / or the like.The connection between each of the plurality of wireless nodes 14 and the core network infrastructure is sometimes referred to as a backhaul connection. The core network infrastructure is a central infrastructure that manages and forwards network traffic from the UE across a wide area network (WAN), including, for example, a mobile switching center (MSC), an Internet backbone, and / or the like.

[0038] By way of non-limiting example, the plurality of wireless nodes 14 are gNodeB cellular base stations (next-generation Node B cellular base stations) configured to provide access to a 5G cellular network. It should be appreciated that the plurality of wireless nodes 14 may be other types of telecommunications nodes, including, for example, a NodeB (i.e., 3G cellular base stations), an eNodeB (i.e., 4G cellular base stations), WiFi access points, and / or the like.

[0039] The plurality of wireless nodes 14 are defined by a plurality of real node parameters. According to an exemplary embodiment, the plurality of real node parameters include, for example, a number of real nodes of the plurality of wireless nodes 14 (e.g., three nodes, as in Fig. 1), a physical node location for each of the plurality of wireless nodes 14 (i.e., a location of each of the plurality of wireless nodes 14 within the environment), a physical node power level for each of the plurality of wireless nodes 14 (i.e., a radio transmission power of each of the plurality of wireless nodes 14, measured, for example, in watts), and a physical node beamforming configuration for each of the plurality of wireless nodes 14 (i.e., operating settings used for beamforming that enable each of the plurality of wireless nodes 14 to focus the transmission or reception of RF signals in a particular direction).

[0040] The plurality of wireless nodes 14 operate within a wireless network environment 26. It should be appreciated that a signal-to-interference and noise ratio (SINR) within the wireless network environment 26 may vary depending on obstacles within the wireless network environment 26. According to the present invention, the system 10 and method 100 are used to estimate the SINR in the wireless network environment 26, as discussed in more detail below.

[0041] Continue in Fig. 1, the plurality of wireless nodes 14 are in electrical communication with the central computer communication system 18 of the one or more central computers 12. Accordingly, the one or more central computers 12 may use the central computer communication system 18 to adjust one or more of the plurality of real node parameters, such as the real node power level and / or the real node beamforming configuration of each of the plurality of wireless nodes 14. Furthermore, the one or more central computers 12 may use the central computer communication system 18 to receive status information from the plurality of wireless nodes 14, as discussed in more detail below.

[0042] The plurality of wireless nodes 14 are also in electrical communication with a plurality of vehicles 28. The plurality of vehicles 28 are considered a UE in the wireless network environment 26 that is in communication with the plurality of wireless nodes 14. Each of the plurality of vehicles 28 includes a vehicle communication system 30 that enables communication between the vehicle systems, such as a vehicle controller, and the plurality of wireless nodes 14. According to an exemplary embodiment, the plurality of vehicles 28 are distributed throughout the wireless network environment 26 and may therefore experience a different SINR. The number of the plurality of vehicles 28 and the volume and type of network traffic transmitted and / or received by each of the plurality of vehicles 28 affect the ambient traffic density of the wireless network environment 26.

[0043] In Fig. 2 shows a flowchart of the method 100 for estimating the signal-to-interference and noise ratio (SINR) in a wireless network environment. The method 100 begins at block 102 and proceeds to block 104. At block 104, the one or more central computers 12 collect a plurality of wireless connection records via the wireless network environment 26 from the UE (i.e., the plurality of vehicles 28). According to an exemplary embodiment, the one or more central computers 12 use the plurality of wireless nodes 14 to collect the plurality of wireless connection records. According to one non-limiting example, each of the plurality of vehicles 28 transmits at least one wireless connection record to the plurality of wireless nodes 14.The plurality of wireless nodes 14 then transmit the plurality of wireless connection records to the one or more central computers 12 via the central computer communication system 18. The one or more central computers 12 store the plurality of wireless connection records in the database 20.

[0044] A wireless connection record includes at least one measured environmental parameter and a measured SINR. The at least one measured environmental parameter includes at least one of: a channel quality indication (CQI) value, a weather condition value, an environmental complexity value, and a communication distance. The CQI value quantifies a level of interference, noise, and signal quality for a wireless connection between one of the plurality of vehicles 28 and the plurality of wireless nodes 14. The weather condition value quantifies a weather condition (e.g., temperature, humidity, precipitation, storm, fog, haze, and / or the like) in the vicinity of one of the plurality of vehicles 28. The environmental complexity value quantifies at least one of: a number of obstacles and an average height of the obstacles that impede wireless transmissions in the vicinity of one of the plurality of vehicles 28.Obstacles that impede wireless transmissions may include, for example, buildings, walls, metal structures, trees, dense foliage, natural terrain (e.g., hills, valleys, etc.), and / or the like. The communication distance quantifies a distance between one of the plurality of vehicles 28 and one or more of the plurality of wireless nodes 14. The measured SINR is a signal-to-interference and noise ratio (SINR) measured by one of the plurality of vehicles 28.

[0045] According to an exemplary embodiment, each of the plurality of vehicles 28 uses a machine learning model for identifying environmental complexity to determine the environmental complexity value. As a non-limiting example, the machine learning model for identifying environmental complexity includes multiple layers, including an input layer and an output layer, as well as one or more hidden layers. The input layer receives environmental data (e.g., geographic location data, elevation data, map data, construction data, property zoning data, a number of obstacles and an average height of the obstacles that impede wireless transmissions, and / or the like) as an input. The input is then passed to the hidden layers. Each hidden layer applies a transformation (e.g.,The output layer applies a nonlinear transformation (a nonlinear transformation) to the data and passes the result to the next hidden layer up to the last hidden layer. The output layer generates the environment complexity value.

[0046] To train the machine learning model for identifying environmental complexity, a dataset of inputs and their corresponding environmental complexity values ​​is used. The algorithm is trained by adjusting internal weights between the nodes in each hidden layer to minimize the prediction error. During training, an optimization technique (e.g., a gradient descent method) is used to adjust the internal weights to reduce the prediction error. The training process is repeated with the entire dataset until the prediction error is minimized. The resulting trained model is then used to classify new input data.

[0047] After sufficient training of the machine learning model to identify the environment complexity, the algorithm can accurately and precisely determine the environment complexity value based on the environment data. By adjusting the weights between the nodes in each hidden layer during training, the algorithm "learns" to recognize patterns in the environment data that indicate the environment complexity value. After block 104, the method 100 proceeds to block 106.

[0048] In block 106, the one or more central computers 12 generate a regression model based at least in part on the plurality of wireless connection data sets collected in block 104. The regression model is configured to determine an estimated SINR based on the at least one measured environmental parameter of each of the plurality of wireless connection data sets collected in block 104. According to one non-limiting example, the regression model is generated using a preprocessing step, an algorithm selection step, a data splitting step, and a training step.

[0049] In the preprocessing step, the one or more central computers 12 cleans the multiple wireless connection data sets collected in block 104. As a non-limiting example, the preprocessing step includes normalization and data cleaning to address missing values ​​and outliers.

[0050] In the algorithm selection step, the one or more central computers 12 select a suitable regression algorithm based on the plurality of wireless connection data sets. By way of non-limiting example, the regression algorithm may include linear regression, logistic regression, comb regression, lasso regression, polynomial regression, or machine learning-based models such as multiple regression or support vector regression. According to an exemplary embodiment, the regression algorithm is selected based on the characteristics of the plurality of wireless connection data sets. For example, linear regression may be selected if there is a linear relationship between the at least one measured environmental parameter and the measured SINR.If the relationship between the at least one measured environmental parameter and the measured SINR is nonlinear, polynomial regression can be chosen.

[0051] In the data splitting step, the multiple wireless connection datasets are split into two parts: a training set and a test set. The training set is used to train the regression model, while the test set is used to evaluate the regression model's performance, as discussed in more detail below. As a non-limiting example, the multiple wireless connection datasets are randomly split to ensure that both the training set and the test set are representative of the entire dataset.

[0052] In the training step, the one or more central computers 12 train the regression algorithm selected during the algorithm selection step. According to an exemplary embodiment, the parameters of the selected regression algorithm are adjusted to minimize a difference between the estimated SINR determined by the regression model based on the training set and the measured SINR in the training set. According to a non-limiting example, the parameters are adjusted using iterative optimization, a gradient descent method, and / or additional optimization techniques. The training step continues until the regression algorithm reaches a predetermined level of accuracy and / or continues until convergence.

[0053] After completing the training step, the regression model can be used to determine the estimated SINR based on the measured environmental parameters. After block 106, the method 100 proceeds to blocks 108 and 110.

[0054] At block 108, the one or more central computers 12 execute the regression model generated at block 106 using the at least one measured environmental parameter of each of the plurality of wireless link data sets in the test set to determine one or more estimated SINRs. As a non-limiting example, the one or more estimated SINRs are stored in database 20. After block 108, the method 100 proceeds to block 112.

[0055] In block 112, the one or more central computers 12 determine a plurality of tested SINR deviations between each of the one or more estimated SINRs determined in block 108 and the measured SINR in each of the plurality of wireless link data sets in the test set. As a non-limiting example, the plurality of tested SINR deviations are stored in database 20. After block 112, the method proceeds to block 114, as discussed in more detail below.

[0056] In Fig. 3 shows a schematic diagram of a simulated wireless network 40 for computer simulation. The simulated wireless network 40 includes a simulated wireless network environment 42 and a plurality of simulated wireless nodes 44. The simulated wireless network environment 42 is defined by at least one parameter of the simulated environment. According to an exemplary embodiment, the at least one parameter of the simulated environment includes, for example, a simulated environment size (e.g., one hundred square kilometers), a simulated ambient traffic density, and one or more simulated obstacles 46. The simulated ambient traffic density models a number of a plurality of simulated vehicles 48 in the simulated wireless network environment 42 and a volume of wireless traffic of the plurality of simulated vehicles 48 in the simulated wireless network environment 42.Each of the plurality of simulated vehicles 48 includes a wireless vehicle communication system 50 that is in wireless electrical communication with one or more of the plurality of simulated wireless nodes 44. The one or more simulated obstacles 46 model obstacles that impede wireless transmissions (e.g., buildings, public infrastructure, and / or the like).

[0057] The plurality of simulated wireless nodes 44 model wireless telecommunications nodes. A wireless node is a location that includes electronic communication devices (e.g., antennas, transceivers, digital signal processors, control electronics, and / or the like) configured to connect the user equipment (UE) to a wireless network (e.g., a cellular network), as discussed above.

[0058] The plurality of simulated wireless nodes 44 are defined by a plurality of simulated node parameters. According to an exemplary embodiment, the plurality of simulated node parameters include, for example, a number of the plurality of simulated wireless nodes 44 (e.g., three nodes, as shown in Fig. 3), a simulated node location for each of the plurality of simulated wireless nodes 44 (i.e., a location of each of the plurality of simulated wireless nodes 44 within the simulated wireless network environment 42), and a simulated node power level for each of the plurality of simulated wireless nodes 44 (i.e., a radio transmission power of each of the plurality of simulated wireless nodes, measured, for example, in watts). The simulated wireless network environment 42 including the plurality of simulated wireless nodes 44 is referred to as the simulated wireless network 40.

[0059] Again in Fig. 2 and continue in Fig. 3, in block 110, the one or more central computers 12 generate a plurality of simulated wireless network environments 42 using computer simulation. To generate one of the plurality of simulated wireless network environments 42, the one or more central computers 12 first initialize the simulated wireless network environment 42. As discussed above, the simulated wireless network environment 42 is defined by the at least one simulated environment parameter.

[0060] According to an exemplary embodiment, the at least one simulated environment parameter defining the simulated wireless network environment 42 is selected by sampling multiple probability distributions corresponding to each of the multiple environmental parameters. The multiple probability distributions include, for example, an environmental size probability distribution, an environmental traffic density probability distribution, and an obstacle probability distribution. According to one non-limiting example, the one or more central computers 12 use Monte Carlo sampling techniques to ensure that such corner and edge cases are accounted for.

[0061] After initializing the simulated wireless network environment 42, the one or more central computers 12 add the plurality of simulated wireless nodes 44 to the simulated wireless network environment 42. According to an exemplary embodiment, the plurality of simulated node parameters are selected by sampling a plurality of probability distributions corresponding to each of the plurality of node parameters. The plurality of probability distributions include, for example, a node count probability distribution, a node location probability distribution, and a node power level probability distribution. According to one non-limiting example, the one or more central computers 12 use Monte Carlo sampling techniques to ensure that such corner and edge cases are accounted for.

[0062] After adding the plurality of simulated wireless nodes 44 to the simulated wireless network environment 42, the one or more central computers 12 measure one or more simulated SINRs within the simulated wireless network environment 42. According to an exemplary embodiment, the one or more central computers 12 simulate the network traffic between the plurality of simulated vehicles 48 and the plurality of simulated wireless nodes 44 to measure the one or more simulated SINRs. During the network traffic simulation, the one or more central computers 12 measure one or more simulated SINRs within the simulated wireless network environment 42.As one non-limiting example, the one or more central computers 12 may perform a simulated video upload task in which each of the plurality of simulated vehicles 48 attempts to upload a high-resolution video to the plurality of simulated wireless nodes 44. The one or more central computers 12 evaluate the one or more simulated SINRs based on a simulated interaction between the simulated network transmissions and the one or more simulated obstacles 46 in the simulated wireless network environment 42. It should be appreciated that the above process for generating one of the plurality of simulated wireless network environments 42 is repeated multiple times to generate the plurality of simulated wireless network environments 42. The at least one parameter of the simulated environment and the plurality of simulated node parameters are varied to provide coverage of corner and / or edge cases.After block 110, the method 100 proceeds to block 116.

[0063] In block 116, the one or more central computers 12 execute the regression model generated in block 106 for each of the plurality of simulated wireless network environments 42 using at least one parameter of the simulated environment of each of the plurality of simulated wireless network environments 42 to determine one or more estimated SINRs for each of the plurality of simulated wireless network environments 42. As one non-limiting example, the one or more estimated SINRs are stored in the database 20 with the one or more estimated SINRs determined in block 108. After block 116, the method 100 proceeds to block 118.

[0064] In block 118, the one or more central computers 12 determine a plurality of simulated SINR deviations by comparing the one or more estimated SINRs for each of the plurality of simulated wireless network environments 42 determined in block 116 and the one or more simulated SINRs measured within the simulated wireless network environment 42 in block 110. As a non-limiting example, the plurality of simulated SINR deviations are stored in the database 20 along with the plurality of tested SINR deviations determined in block 112. After block 118, the method proceeds to block 114.

[0065] In block 114, the one or more central computers 12 train a machine learning algorithm for SINR optimization. The machine learning algorithm is used for SINR optimization to increase the accuracy of the SINR estimation by compensating for the errors introduced by the regression model and generating an optimized estimated SINR as an output. According to one non-limiting example, the machine learning algorithm for SINR optimization is trained to account for edge and / or corner cases that may be underrepresented in the multiple wireless link datasets.As one non-limiting example, the SINR optimization machine learning algorithm is trained based on the one or more estimated SINRs of each of the plurality of wireless link data sets in the test set, the plurality of tested SINR deviations determined in block 112, the one or more estimated SINRs for each of the plurality of simulated wireless network environments 42, the plurality of simulated SINR deviations determined in block 118, the at least one measured environmental parameter of each of the plurality of wireless link data sets, the at least one simulated environment parameter of each of the plurality of simulated wireless network environments, the measured SINR of each of the plurality of wireless data sets, and the simulated SINR of each of the plurality of simulated wireless network environments.

[0066] As one non-limiting example, the SINR optimization machine learning algorithm includes multiple layers, including an input layer and an output layer, as well as one or more hidden layers. The input layer receives as inputs the one or more estimated SINRs from each of the plurality of wireless link data sets in the test set, the plurality of tested SINR deviations determined in block 112, the one or more estimated SINRs for each of the plurality of simulated wireless network environments 42, the plurality of simulated SINR deviations determined in block 118, the at least one measured environmental parameter of each of the plurality of wireless link data sets, and the at least one simulated environmental parameter of each of the plurality of simulated wireless network environments. The inputs are then passed to the hidden layers.Each hidden layer applies a transformation (e.g., a nonlinear transformation) to the data and passes the result to the next hidden layer until the last hidden layer. The output layer produces the optimized estimated SINR.

[0067] To train the machine learning algorithm for SINR optimization, a dataset of inputs and their corresponding optimized estimated SINR (i.e., the measured SINR of each of the multiple wireless datasets and the simulated SINR of each of the multiple simulated wireless network environments) is used. The algorithm is trained by adjusting the internal weights between nodes in each hidden layer to minimize the prediction error. During training, an optimization technique (e.g., a gradient descent method) is used to adjust the internal weights to reduce the prediction error. The training process is repeated with the entire dataset until the prediction error is minimized, with the resulting trained algorithm then used to generate the optimized estimated SINR based on new input data.

[0068] After sufficient training of the SINR optimization machine learning algorithm, the algorithm can accurately and precisely determine the optimized estimated SINR based on the plurality of tested SINR deviations, the plurality of simulated SINR deviations, the at least one measured environmental parameter of each of the plurality of wireless link data sets, the at least one simulated environmental parameter of each of the plurality of simulated wireless network environments, the measured SINR of each of the plurality of wireless data sets, and the simulated SINR of each of the plurality of simulated wireless network environments. By adjusting the weights between the nodes in each hidden layer during training, the algorithm "learns" to recognize patterns in the data that indicate the optimized estimated SINR. After block 114, the method 100 proceeds to enter a standby state at block 120.

[0069] According to an exemplary embodiment, the one or more central computers 12 repeatedly exit the standby state 120, restarting the method 100 at block 102. According to a non-limiting example, the one or more central computers 12 exit the standby state 120, restarting the method 100 via a timer, e.g., every three hundred milliseconds.

[0070] In Fig.4 shows a flowchart of a method 400 for using the regression model generated in block 106 and the machine learning algorithm trained in block 114 for SINR optimization. According to one exemplary embodiment, the method 400 is executed by the one or more central computers 12. According to another exemplary embodiment, the method 400 is executed by another computer system, such as the vehicle controller of one or more of the plurality of vehicles 28. The method 400 begins in block 402 and proceeds to block 404. In block 404, at least one measured environmental parameter is collected.

[0071] According to an exemplary embodiment, the at least one measured environmental parameter is determined using sensors, including, for example, vehicle sensors configured to measure at least one of: a channel quality indicator (CQI) value, a weather condition value, an environmental complexity value, and a communication distance. After block 404, the method 400 proceeds to block 406.

[0072] In block 406, the regression model generated in block 106 is executed using the at least one measured environmental parameter collected in block 404 to determine an estimated SINR. After block 406, the method 400 proceeds to block 408.

[0073] At block 408, the SINR optimization machine learning algorithm is executed using the at least one measured environmental parameter and the estimated SINR as input to generate an optimized estimated SINR as output. After block 408, the method 400 proceeds to block 410.

[0074] At block 410, the optimized estimated SINR is used to improve the operation of a wireless communication system. According to an exemplary embodiment, the wireless communication system is vehicle communication system 30. According to one non-limiting example, transmission parameters, such as transmit power, beamforming parameters, and / or the like, may be adjusted based on the optimized estimated SINR. According to another non-limiting example, the vehicle controller may delay or accelerate the transmission of critical data based on the optimized estimated SINR.

[0075] According to another exemplary embodiment, the optimized estimated SINR is transmitted to a central computing system (e.g., the one or more central computers 12) and stored (e.g., in the database 20) for analysis of network coverage and performance within the wireless network environment 26. The optimized estimated SINR values ​​can be analyzed to support the development and design of a wireless communications infrastructure. After block 410, the method 400 continues to enter a standby state at block 412.

[0076] According to an exemplary embodiment, the method 400 repeatedly exits the standby state 412 and restarts at block 402. As a non-limiting example, the method 400 exits the standby state 412 and restarts via a timer, e.g., every three hundred milliseconds.

[0077] The system 10 and methods 100, 400 of the present invention provide several advantages. By training the regression model based on multiple wireless link datasets, the regression model can determine the estimated SINR for each location in the environment based on measured environmental parameters, thereby reducing the reliance on low-resolution SINR estimation techniques. Training the machine learning SINR optimization algorithm enables optimization of the estimated SINR based on edge and / or corner cases generated using a computer simulation. Method 400 demonstrates an exemplary practical application of the system 10 and method 100 to determine optimized estimated SINRs.

Claims

[1] A method for estimating the signal-to-interference and noise ratio (SINR) in a wireless network environment, the method comprising: Collecting a plurality of wireless connection data sets about the environment of a wireless network, each of the plurality of wireless connection data sets including at least one measured environmental parameter and a measured SINR; Generating a regression model based at least in part on the plurality of wireless connection data sets, the regression model configured to determine an estimated SINR based on the at least one measured environmental parameter; and training a SINR optimization machine learning algorithm to determine an optimized estimated SINR based at least in part on the estimated SINR; characterized by , that collecting the plurality of wireless connection data sets further comprises: Receiving the plurality of wireless connection data sets from one or more wireless devices, wherein the at least one measured environmental parameter of each of the plurality of wireless connection data sets includes at least one of a channel quality indication (CQI) value, a weather condition value, and an environmental complexity value; wherein the environmental complexity value quantifies a number of obstacles and / or an average height of the obstacles that hinder wireless transmissions. [2] The method of claim 1, wherein the environmental complexity value is determined by each of the one or more wireless devices using a machine learning model to identify environmental complexity. [3] The method of claim 1, wherein the one or more wireless devices include at least one vehicle equipped with a vehicle communication system. [4] The method of claim 1, wherein training the machine learning algorithm for SINR optimization further comprises: Determining a plurality of tested SINR deviations between the estimated SINR determined by the regression model and the measured SINR in each of the plurality of wireless connection data sets; and Training the SINR optimization machine learning algorithm based at least in part on the plurality of tested SINR deviations, wherein the SINR optimization machine learning algorithm is trained to receive the at least one measured environmental parameter and the estimated SINR as an input and to provide the optimized estimated SINR as an output. [5] The method of claim 4, wherein training the machine learning algorithm for SINR optimization further comprises: Training the SINR optimization machine learning algorithm to consider edge cases using multiple simulated wireless network environments. [6] The method of claim 5, wherein training the SINR optimization machine learning algorithm to account for edge cases using multiple simulated wireless network environments further comprises: Creating multiple simulated wireless network environments using a computer simulation; and Training the machine learning algorithm for SINR optimization based at least in part on the multiple simulated wireless network environments. [7] The method of claim 6, wherein generating the plurality of simulated wireless network environments further comprises: Initializing the plurality of simulated wireless network environments, wherein each of the plurality of simulated wireless network environments is defined by at least one parameter of the simulated environment; adding a plurality of simulated wireless nodes to each of the plurality of simulated wireless network environments to form a plurality of simulated wireless network environments; and Measure one or more simulated SINRs in each of several simulated wireless network environments. [8] The method of claim 7, wherein training the machine learning algorithm for SINR optimization based at least in part on the plurality of simulated wireless network environments further comprises: Determining one or more estimated SINRs for each of the plurality of simulated wireless network environments using the regression model based at least in part on the at least one simulated environment parameter of each of the plurality of simulated wireless network environments; Comparing the one or more estimated SINRs with the one or more simulated SINRs for each of the plurality of simulated wireless network environments to determine a plurality of simulated SINR deviations; and Training the SINR optimization machine learning algorithm based at least in part on the at least one simulated environment parameter of each of the plurality of simulated wireless network environments and the plurality of simulated SINR deviations.

Images