Wireless communication quality prediction device, wireless communication quality prediction method, and wireless communication quality prediction program
The wireless communication quality prediction device addresses the inflexibility and accuracy issues of existing methods by generating and synthesizing prediction models for multiple spatial information types, ensuring accurate and adaptable communication quality forecasting.
Patent Information
- Authority / Receiving Office
- JP · JP
- Patent Type
- Patents
- Current Assignee / Owner
- Filing Date
- 2022-08-22
- Publication Date
- 2026-03-26
AI Technical Summary
Existing wireless communication quality prediction methods struggle to flexibly update prediction models in response to changes in data collection equipment and lack accuracy when using multiple types of physical spatial information.
A wireless communication quality prediction device that generates prediction models for each type of physical spatial information, calculates reliability scores, and synthesizes these models using maximum ratio synthesis to predict future communication quality accurately.
Enables flexible updates to prediction models and achieves higher accuracy in predicting wireless communication quality compared to conventional methods, while detecting potential failures in data collection devices.
Smart Images
Figure 0007836001000002 
Figure 0007836001000003 
Figure 0007836001000004
Abstract
Description
Technical Field
[0001] The present invention relates to a wireless communication quality prediction device, a wireless communication quality prediction method, and a wireless communication quality prediction program.
Background Art
[0002] When using a communication device equipped with a wireless communication function, the wireless communication quality may change with changes in the surrounding environment of the communication device (such as the movement of objects existing around the communication device). Such changes in wireless communication quality become factors that prevent the communication device from enjoying services or the wireless communication quality required by the service providing system from being satisfied. Therefore, there is a technology for predicting the wireless communication quality of a communication device using a prediction model of wireless communication quality (Non-Patent Documents 1-3).
Prior Art Documents
Non-Patent Documents
[0003]
Non-Patent Document 1
Non-Patent Document 2
Non-Patent Document 3
[0004] Non-patent documents 1 and 2 predict the quality of wireless communication when the wireless communication path is obstructed by the passage of an object, using physical spatial information acquired from a depth camera. However, non-patent documents 1 and 2 only predict wireless communication quality using a single piece of physical spatial information acquired from a depth camera, and cannot predict it using multiple types of physical spatial information.
[0005] Non-patent document 3 predicts wireless communication quality using a multimodal learning method that takes multiple types of physical spatial information as input. However, with Non-patent document 3, it is difficult to flexibly update the prediction model in response to updates to the collection equipment used to collect physical spatial information (such as adding, removing, or moving cameras).
[0006] This invention has been made in view of the above circumstances, and aims to provide a technology that can flexibly update the prediction model for wireless communication quality in response to updates to collection devices for collecting physical spatial information, and that can predict future wireless communication quality with higher accuracy than conventional multimodal learning methods. [Means for solving the problem]
[0007] A wireless communication quality prediction device according to one aspect of the present invention comprises: a generation unit that generates a prediction model for each piece of physical space information, based on wireless communication information of a communication device and a plurality of different pieces of physical space information about the vicinity of the communication device, and outputs a prediction value of the future wireless communication quality of the communication device, using the physical space information as input; a generation unit that generates a reliability model for each piece of physical space information, using the prediction value output from the prediction model, and outputs the reliability of the prediction value; a calculation unit that calculates a plurality of prediction values of the future wireless communication quality of the communication device to be predicted by inputting a plurality of different pieces of physical space information about the vicinity of the communication device to be predicted into a plurality of prediction models corresponding to the plurality of physical space information; a calculation unit that calculates the reliability of each of the plurality of prediction values by inputting a plurality of different pieces of physical space information about the vicinity of the communication device to be predicted into a plurality of reliability models corresponding to the plurality of physical space information; and a synthesis unit that synthesizes the plurality of prediction values by maximum ratio synthesis based on the reliability of each of the plurality of prediction values.
[0008] A wireless communication quality prediction method according to one aspect of the present invention is a wireless communication quality prediction method performed by a wireless communication quality prediction device, comprising the steps of: generating a prediction model that outputs a predicted value of the future wireless communication quality of a communication device based on wireless communication information of a communication device and a plurality of different physical space information of the surrounding area of the communication device, with the physical space information as input for each physical space information; generating a reliability model that outputs the reliability of the predicted value using the predicted value output from the prediction model with the physical space information as input for each physical space information; calculating a plurality of predicted values of the future wireless communication quality of a communication device to be predicted by inputting a plurality of different physical space information of the surrounding area of the communication device to be predicted into a plurality of prediction models corresponding to the plurality of physical space information; calculating the reliability of each of the plurality of predicted values by inputting a plurality of different physical space information of the surrounding area of the communication device to be predicted into a plurality of reliability models corresponding to the plurality of physical space information; and synthesizing the plurality of predicted values by maximum ratio synthesis based on the reliability of each of the plurality of predicted values.
[0009] One embodiment of the present invention is a wireless communication quality prediction program that causes a computer to function as the wireless communication quality prediction device. [Effects of the Invention]
[0010] According to the present invention, it is possible to flexibly update the prediction model for wireless communication quality in response to updates to the collection device for collecting physical spatial information, and to provide a technology that can predict future wireless communication quality with higher accuracy than conventional multimodal learning methods. [Brief explanation of the drawing]
[0011] [Figure 1] Figure 1 shows an example of the overall configuration of a wireless communication quality prediction system. [Figure 2] Figure 2 shows the processing flow during the preparation phase. [Figure 3]Figure 3 shows examples of physical spatial information, predictive models, and confidence models. [Figure 4] Figure 4 shows the processing flow during the prediction stage. [Figure 5] Figure 5 shows an example of the hardware configuration of a wireless communication quality prediction device. [Modes for carrying out the invention]
[0012] Embodiments of the present invention will be described below with reference to the drawings. In the drawings, identical parts are denoted by the same reference numerals and their descriptions are omitted.
[0013] [Overview of the prefecture] In the future, cameras and sensors are expected to become widespread in urban areas. Therefore, the present invention aims to provide a technology that can flexibly update the prediction model for wireless communication quality in response to updates (addition, reduction, relocation, etc.) of collection devices for collecting physical spatial information, such as cameras and sensors, and that can predict future wireless communication quality with higher accuracy compared to conventional multimodal learning methods.
[0014] To achieve this objective, the present invention generates a predictive model for wireless communication quality for each of several different physical spatial information sources. Furthermore, the present invention generates a reliability model for calculating the reliability of wireless communication quality and synthesizes multiple wireless communication quality sources by combining the maximum ratio based on the reliability of each of the multiple wireless communication quality sources.
[0015] Thus, since the present invention generates a prediction model for wireless communication quality for each of several different pieces of physical spatial information, the prediction model for wireless communication quality can be updated in a short time when the data acquisition device is updated. Furthermore, it eliminates the need to centrally manage the various data collected by the data acquisition device.
[0016] In addition, the present invention generates a reliability model for calculating the reliability of wireless communication quality, and synthesizes a plurality of wireless communication qualities by maximum ratio combining based on the reliability of each of the plurality of wireless communication qualities. Therefore, it is possible to achieve prediction accuracy equal to or higher than that of a multimodal learning method that takes a plurality of types of physical space information as input. In addition, since the reliability of wireless communication quality is calculated, it is possible to detect a failure of the collection device.
[0017] [Configuration of Wireless Communication Quality Prediction System] FIG. 1 is a diagram showing an overall configuration example of a wireless communication quality prediction system according to the present embodiment. The wireless communication quality prediction system includes a wireless communication quality prediction device 1 and a plurality of collection devices 2.
[0018] The collection device 2 is a camera or a sensor installed in an urban area or the like, and can be updated at any time according to various situations. The collection device 2 collects physical space information around the communication device 100 and returns the physical space information in response to an acquisition request from the wireless communication quality prediction device 1.
[0019] Here, the physical space information will be described. The physical space information is, for example, the bounding box coordinate information of an object O (such as a pedestrian or a vehicle) photographed by a camera, the coordinate information of the skeleton of an object O (a pedestrian) photographed by a camera, and the GPS information of the communication device 100. One collection device 2 may collect one piece of physical space information or a plurality of pieces of physical space information.
[0020] The wireless communication quality prediction device 1 generates a prediction model for predicting the wireless communication quality of the communication device 100 based on the physical space information around the communication device 100 and the wireless communication information of the communication device 100, and further generates a reliability model for calculating the reliability of the prediction model. Using the prediction model and the reliability model, it predicts the future wireless communication quality of the communication device 100 (here, the communication device to be predicted) communicating with the communication device 200 of the communication partner via the Internet N or the like.
[0021] The wireless device being predicted may communicate wirelessly with one communication device, or with two or more communication devices. Wireless communication is not limited to communication between two communication devices; it may also be wireless communication between three or more communication devices.
[0022] [Functional Configuration of Wireless Communication Quality Prediction Device] Next, the functional configuration of the wireless communication quality prediction device 1 will be described. As shown in Figure 1, the wireless communication quality prediction device 1 comprises a wireless communication unit 11, a plurality of physical space information acquisition units 12, a data storage unit 13, a prediction model generation unit 14, a reliability model generation unit 15, a communication quality calculation unit 16, a reliability calculation unit 17, and a communication quality synthesis unit 18.
[0023] The wireless communication unit 11 has the function of communicating wirelessly with the communication device 100, acquiring wireless communication information from the communication device 100, and storing it in the data storage unit 13. Wireless communication information includes, for example, the radio wave strength, communication speed, and location information of the communication device 100. Wireless communication information is not limited to wireless communication quality information relating to the quality of wireless communication, but also includes wireless communication auxiliary information to support wireless communication quality information.
[0024] Multiple physical space information acquisition units 12 correspond to multiple collection devices 2 and have the function of acquiring multiple different physical space information about the vicinity of the communication device 100 from the multiple collection devices 2 and storing it in the data storage unit 13. If one collection device 2 is capable of collecting multiple physical space information, then only one physical space information acquisition unit 12 is required.
[0025] The data storage unit 13 has the function of storing, in association with, the wireless communication information of the communication device 100 acquired by the wireless communication unit 11 and the multiple physical space information acquisition units 12 each acquired at the same time as the acquisition of the wireless communication information.
[0026] Specifically, the data storage unit 13 classifies multiple pieces of physical space information into categories, associates each piece of physical space information with wireless communication information from the communication device 100 acquired at the same time, and generates a database for inputting physical space information of the category corresponding to the category of the prediction model generated by the prediction model generation unit 14.
[0027] Here, we will explain the categories. A category is a classification based on the type of physical spatial information. For example, there are categories such as "bounding box coordinate information of pedestrians," "bounding box coordinate information of vehicles," "coordinate information of pedestrian skeletons," "GPS information of communication device 100," and "communication quality." Note that physical spatial information also differs depending on the type and location of collection device 2, so the type and location of collection device 2 may also be used as categories.
[0028] The prediction model generation unit 14 has the function of generating a prediction model that takes the physical space information as input and outputs a predicted value for the future wireless communication quality of the communication device 100, based on the wireless communication information of the communication device 100 and multiple different pieces of physical space information around the communication device 100.
[0029] Specifically, the prediction model generation unit 14 uses the database stored in the data storage unit 13 (a data set that correlates physical space information and wireless communication information for each category) to generate a prediction model for each category to predict the wireless communication quality of the communication device 100.
[0030] The confidence model generation unit 15 has a function to generate a confidence model that outputs the confidence level of the predicted values using the predicted values output from the prediction model, taking the physical spatial information as input, for each piece of physical spatial information.
[0031] Specifically, the confidence model generation unit 15 generates a confidence model for calculating the confidence score of the prediction model. The confidence score of the prediction model is a numerical value that quantifies the accuracy of the prediction model, and is, for example, the absolute error between the predicted value and the observed value of wireless communication quality (|(predicted value - observed value)|) and its absolute percentage error (|(predicted value - observed value) / observed value|). The confidence score of the prediction model is an example of the confidence level of the predicted value.
[0032] Furthermore, since both the prediction model generation unit 14 and the confidence model generation unit 15 have common functions for generating learning models such as prediction models and confidence models, they may be implemented in a single generation unit. In this case, the single generation unit may generate a learning model that outputs both a predicted value of wireless communication quality and a confidence score of the predicted value.
[0033] The communication quality calculation unit 16 has the function of calculating multiple predicted values for the future wireless communication quality of the communication device 100 (the communication device to be predicted) by inputting multiple different pieces of physical spatial information about the surrounding area of the communication device 100 (the communication device to be predicted) into multiple prediction models corresponding to the multiple pieces of physical spatial information.
[0034] Specifically, the communication quality calculation unit 16 acquires multiple physical spatial information at a time after the generation of the prediction model and the reliability model, obtains a prediction model for each category corresponding to each category of the acquired multiple physical spatial information from the prediction model generation unit 14, and calculates multiple prediction values for wireless communication quality by inputting the multiple physical spatial information into each prediction model.
[0035] The reliability calculation unit 17 has the function of calculating the reliability of each of the multiple predicted values predicted by the multiple prediction models by inputting multiple different pieces of physical spatial information about the vicinity of the communication device 100 (the communication device to be predicted) into multiple reliability models corresponding to the multiple pieces of physical spatial information.
[0036] Specifically, the confidence calculation unit 17 acquires multiple physical spatial information at a time after the generation of the prediction model and the confidence model, obtains the confidence model for each category corresponding to each category of the acquired multiple physical spatial information from the confidence model generation unit 15, and calculates the confidence score for each prediction value by inputting the multiple physical spatial information into each confidence model.
[0037] The communication quality synthesis unit 18 has a function to synthesize multiple predicted values by maximum ratio synthesis based on the confidence level of each of the multiple predicted values. Specifically, the communication quality synthesis unit 18 synthesizes the predicted values of the prediction model by maximum ratio synthesis based on confidence scores, and outputs the synthesized predicted values as the future wireless communication quality of the communication device 100 (the communication device to be predicted).
[0038] [Operation of the Wireless Communication Quality Prediction Device] Next, the operation of the wireless communication quality prediction device 1 will be described.
[0039] The operation of the wireless communication quality prediction device 1 can be broadly divided into two stages. The first is a preparation stage in which a prediction model and a confidence model are generated. The second is a prediction stage, which takes place after the preparation stage and uses the generated prediction model and confidence model to infer the future wireless communication quality of the communication device 100. The prediction stage can be divided into two stages: data acquisition and data processing.
[0040] First, let's explain the operation during the preparation phase. Figure 2 shows the processing flow during the preparation phase.
[0041] Step S101; The wireless communication unit 11 communicates wirelessly with the communication device 100 and acquires wireless communication information from the communication device 100.
[0042] Step S102; Next, at the same time as in step S101, the multiple physical space information acquisition units 12 acquire multiple different physical space information of the surroundings of the communication device 100 from the multiple acquisition devices 2. The acquisition devices 2 may be cameras or sensors installed around the communication device 100, or cameras or sensors provided in the communication device 100.
[0043] Step S103; Next, the data storage unit 13 classifies multiple pieces of physical spatial information into categories and generates a database that associates each piece of physical spatial information with wireless communication information. An example of physical spatial information for each category is shown in Figure 3(a). In this example, multiple different pieces of "bounding box coordinate information," "skeleton (motion) coordinate information," and "GPS information of the communication device" are used as categories of physical spatial information.
[0044] Step S104; Next, the prediction model generation unit 14 uses the physical space information for each category stored in the database and the wireless communication information corresponding to that physical space information to generate a prediction model that takes the physical space information as input and outputs a predicted value for the future wireless communication quality of the communication device 100 for each category (= for each type of physical space information).
[0045] Step S105; Finally, the confidence model generation unit 15 takes the physical spatial information as input for each category (= for each type of physical spatial information) and generates a confidence model that outputs a confidence score for the predicted value of the wireless communication quality, using the predicted value of the wireless communication quality output from the prediction model.
[0046] Figure 3(b) shows examples of prediction models and confidence models for each category. In this example, the prediction model corresponding to the "bounding box coordinate information 1" category is designated as prediction model 1, and the predicted value output by prediction model 1 is denoted as R1 (with a hat symbol above R1). Furthermore, the confidence model corresponding to the same "bounding box coordinate information 1" category is designated as confidence model 1, and the confidence score of the predicted value R1 output by confidence model 1 is denoted as ρ1.
[0047] Subsequently, the wireless communication quality prediction device 1 repeatedly executes steps S101 to S105 at multiple different times. In other words, it repeatedly learns the prediction model and the confidence model. These multiple different times include, for example, regular times, irregular times, and times when the data collection device 2 is updated (addition, reduction, movement, etc.).
[0048] For example, if a predetermined data collection device 2 is added and a new category of physical spatial information is acquired, a new prediction model and confidence model corresponding to that new category are generated. Alternatively, if a predetermined data collection device 2 is removed and it becomes impossible to acquire a predetermined category of physical spatial information, the prediction model and confidence model corresponding to that predetermined category are deleted or not trained.
[0049] Next, we will explain the operation of the prediction phase. Figure 4 is a diagram showing the processing flow of the prediction phase.
[0050] Step S201; First, multiple physical space information acquisition units 12 acquire multiple different pieces of physical space information from multiple collection devices 2 about the area around the communication device 100 (the communication device to be predicted).
[0051] Step S202; Next, the data storage unit 13 classifies the multiple physical space information into categories and stores the multiple physical space information into each category of physical space information.
[0052] Step S203; Next, the communication quality calculation unit 16 acquires a prediction model for each category of the multiple physical spatial information classified in step S202, and calculates multiple prediction values for wireless communication quality by inputting the multiple physical spatial information into each prediction model.
[0053] Next, the confidence calculation unit 17 obtains confidence models for each category of the multiple physical spatial information classified in step S202, and calculates the confidence score for each predicted value by inputting the multiple physical spatial information into each confidence model.
[0054] Step S204; Next, the communication quality synthesis unit 18 selects multiple prediction values with the highest confidence scores from among the multiple prediction values, and synthesizes the selected multiple prediction values by performing maximum ratio synthesis based on the confidence scores of each of the selected multiple prediction values.
[0055] For example, the communication quality synthesis unit 18 synthesizes the predicted values of each prediction model using the confidence score, based on the maximum ratio synthesis calculation formula shown in equation (1).
[0056]
number
[0057] Note, R Ns ρ represents the predicted value for each prediction model. Ns This is the predicted value R of each prediction model. Ns This is the confidence score. s This is the index of the prediction models to be combined. It can be specified appropriately depending on the situation, such as using the index of the top two prediction models based on their confidence scores. Of course, all prediction values can also be specified for combination.
[0058] Step S205; Finally, the communication quality synthesis unit 18 outputs the synthesized predicted value as the future wireless communication quality of the communication device 100 (the communication device to be predicted).
[0059] Up to this point, we have explained the case where the prediction model and the confidence model are separated, but as mentioned above, a learning model that outputs both predicted values and confidence scores is also acceptable. In that case, since there is a difference in the number of digits between the number of digits of the predicted values and the number of digits of the confidence scores, it is advisable to weight the model with the smaller number of digits when calculating the loss function during the training of the learning model.
[0060] [effect] According to this embodiment, in the wireless communication quality prediction device 1, the prediction model generation unit 14 generates a prediction model for wireless communication quality for each of several different physical spatial information sources, so that the prediction model for wireless communication quality can be updated in a short time in response to updates to the data collection device 2. Furthermore, it is not necessary to centrally manage the various data collected by the data collection device 2.
[0061] Furthermore, according to this embodiment, in the wireless communication quality prediction device 1, the reliability model generation unit 15 generates a reliability model for calculating the reliability score of wireless communication quality, the communication quality calculation unit 16 calculates multiple predicted values using prediction models for each piece of physical space information, the reliability calculation unit 17 calculates the reliability score for each of the calculated multiple predicted values, and the communication quality synthesis unit 18 synthesizes multiple wireless communication quality values by maximum ratio synthesis based on each reliability score. As a result, prediction accuracy equivalent to or better than that of a multimodal learning method that takes multiple types of physical space information as input can be achieved. In addition, since the reliability of wireless communication quality is calculated, it becomes possible to detect failures in the data collection device.
[0062] [others] The present invention is not limited to the embodiments described above. Numerous modifications are possible within the scope of the spirit of the present invention.
[0063] The wireless communication quality prediction device 1 of this embodiment described above can be realized using a general-purpose computer system, for example, as shown in Figure 5, which includes a CPU 901, a memory 902, a storage 903, a communication device 904, an input device 905, and an output device 906. The memory 902 and the storage 903 are storage devices. In this computer system, the CPU 901 executes a predetermined program loaded onto the memory 902, thereby realizing each function of the wireless communication quality prediction device 1.
[0064] The wireless communication quality prediction device 1 may be implemented on a single computer. The wireless communication quality prediction device 1 may be implemented on multiple computers. The wireless communication quality prediction device 1 may also be a virtual machine implemented on a computer. The program for the wireless communication quality prediction device 1 can be stored on a computer-readable recording medium such as an HDD, SSD, USB memory, CD, or DVD. The program for the wireless communication quality prediction device 1 can also be distributed via a communication network. [Explanation of Symbols]
[0065] 1: Wireless communication quality prediction device 11: Wireless Communication Department 12: Physical Spatial Information Acquisition Unit 13: Data storage unit 14: Predictive Model Generation Unit 15: Confidence Model Generation Unit 16: Communication Quality Calculation Unit 17: Confidence Calculation Unit 18: Communication Quality Synthesis Unit 2: Collection device 100: Communication device 200: Communication equipment 901:CPU 902: Memory 903: Storage 904: Communication device 905: Input device 906: Output device
Claims
1. A generation unit generates a prediction model that, based on wireless communication information of a communication device and multiple different pieces of physical space information surrounding the communication device, outputs a predicted value of the future wireless communication quality of the communication device, taking the physical space information as input for each piece of physical space information. A generation unit generates a confidence model that outputs the confidence level of the predicted value using the predicted value output from the prediction model, with the physical space information as input for each of the aforementioned physical space information. A calculation unit calculates multiple predicted values for the future wireless communication quality of a communication device by inputting multiple different pieces of physical spatial information about the surrounding area of the communication device to be predicted into multiple prediction models corresponding to the multiple pieces of physical spatial information. A calculation unit calculates the reliability of each of the multiple predicted values by inputting multiple different pieces of physical spatial information about the vicinity of the communication device to be predicted into multiple reliability models corresponding to the multiple pieces of physical spatial information, A synthesis unit that synthesizes the plurality of predicted values by combining the maximum ratio based on the confidence levels of each of the plurality of predicted values, A wireless communication quality prediction device equipped with the following features.
2. In a wireless communication quality prediction method performed by a wireless communication quality prediction device, The steps include generating a predictive model that outputs a predicted value of the future wireless communication quality of the communication device, based on wireless communication information of the communication device and a plurality of different physical space information around the communication device, using the physical space information as input for each physical space information, and For each piece of physical space information, the steps include: generating a confidence model that outputs the confidence level of the predicted value using the predicted value output from the prediction model with the physical space information as input; The steps include: inputting multiple different pieces of physical spatial information about the vicinity of the communication device to be predicted into multiple prediction models corresponding to the multiple pieces of physical spatial information to calculate multiple predicted values for the future wireless communication quality of the communication device to be predicted; The steps include: inputting multiple different pieces of physical spatial information about the vicinity of the communication device to be predicted into multiple confidence models corresponding to the multiple pieces of physical spatial information to calculate the confidence level of each of the multiple predicted values; The steps include: synthesizing the plurality of predicted values by combining the maximum ratio based on the confidence levels of each of the plurality of predicted values; A method for predicting wireless communication quality.
3. A wireless communication quality prediction program that causes a computer to function as a wireless communication quality prediction device according to claim 1.
Citation Information
Patent Citations
Ai-based, semi-supervised interactive map enrichment for radio access network planning
US20220051057A1
Technique for assessing connection quality
WO2022069061A1