Estimation device, learning device, estimation method, learning method, and program

Abstract

An estimation device for estimating a range of a change in communication quality at a certain position, the estimation device comprising an input unit which inputs position information indicating the position and an estimation unit which uses a learned model to estimate one or more statistical values regarding the communication quality from the position information and estimates the range of the change on the basis of the one or more statistical values.

Description

Technical Field

[0001] The present invention relates to a technique for estimating wireless communication quality.

Background Art

[0002] In applications such as remote control or remote monitoring of moving objects such as vehicles and robots, it is common to perform remote communication between a server at a center and a terminal provided on the moving object at the site by wireless communication. In addition, due to the movement of the moving object, high availability of the wireless communication quality of the remote communication is required in consideration of safety. It is considered effective to estimate (predict) whether the wireless communication quality can be stably used at the moving destination of the moving object, and if there is a risk, perform processes such as reducing the moving speed in advance or reducing the video transmission rate so as not to have video interruptions. Therefore, it is important to estimate the communication quality at the moving destination.

[0003] As an example of a method for estimating the wireless communication quality at the moving destination, there is a technique for estimating in advance the communication quality at the moving destination based on a history showing the relationship between the position of the moving object and the communication quality in the past (see Non-Patent Document 1). In the technique disclosed in this Non-Patent Document 1, a technique for estimating the communication quality of throughput as a median value or an average value by machine learning or the like using past measured values (actual values) of communication quality has been proposed.

Prior Art Documents

Non-Patent Documents

[0004]

Non-Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] However, in wireless communication of moving objects whose location changes, there can be significant fluctuations in the radio wave environment and communication quality. Therefore, conventional estimates of communication quality as median or average values ​​cannot accurately determine whether the desired communication quality can be continuously met.

[0006] This invention has been made in view of the above points, and aims to estimate, with the utmost accuracy, whether a desired communication quality can be continuously met, even in the case of wireless communication by a moving object whose position is changing. [Means for solving the problem]

[0007] According to the disclosed technology, an estimation device for estimating the range of changes in communication quality at a certain location, An input unit for inputting position information indicating the aforementioned position, Using a trained model, one or more statistical values ​​relating to the communication quality are obtained from the location information. Mean and variance We estimate, In a normal distribution having the mean and variance mentioned above, the upper and lower limits of the range in which the probability of a value falling within that range is the desired distribution range P[%] are Range of the aforementioned change as The estimation part that estimates An estimation device is provided that includes the following. [Effects of the Invention]

[0008] According to the disclosed technology, even in the case of wireless communication by a moving object whose position is changing, it becomes possible to estimate with the greatest possible accuracy whether the desired communication quality can be continuously met. [Brief explanation of the drawing]

[0009] [Figure 1] This is a diagram showing the configuration of the learning device 100 in an embodiment of the present invention. [Figure 2] This is a flowchart illustrating the operation of the learning device 100. [Figure 3] This is a diagram showing the configuration of the estimation device 200 in an embodiment of the present invention. [Figure 4]It is a diagram showing a configuration example of the estimation unit 220. [Figure 5] It is a flowchart for explaining the operation of the estimation device 200. [Figure 6] It is a diagram showing examples of the upper limit value and the lower limit value. [Figure 7] It is a diagram showing a hardware configuration example of the device.

Embodiments for Carrying Out the Invention

[0010] Hereinafter, embodiments (these embodiments) of the present invention will be described with reference to the drawings. The embodiments described below are merely examples, and the embodiments to which the present invention is applied are not limited to the following embodiments.

[0011] In the following description, it is assumed that "communication quality" is the quality in wireless communication. However, the technology according to the present invention is also applicable to the communication quality of communication performed by wire instead of wireless. Also, in the following description, unless otherwise specified, it is assumed that the communication quality is the quality of that communication when the terminal performs wireless communication. Also, "communication quality" may be referred to as "quality".

[0012] It is assumed that the "terminal" moves. The "terminal" may be a smartphone or the like held by a person, or a communication device mounted on a moving body such as a drone, an automobile, or a robot, or the moving body itself such as a drone, an automobile, or a robot may be referred to as the "terminal".

[0013] Also, in the following description, the learning device 100 and the estimation device 200 will be described respectively, but the learning device 100 and the estimation device 200 may be the same device.

[0014] Note that the "communication quality" used below may be any of received power, throughput, delay, jitter, and packet loss, or the "communication quality" may be a quality other than these. Also, the "model" described below may be referred to as a "learning device".

[0015] In the following description, a normal distribution is used as the distribution of communication quality assumed when estimating the range of changes in communication quality. However, this is just an example. Other distributions may be used as the distribution of communication quality assumed when estimating the range of changes in communication quality.

[0016] (Overview of the Embodiment) In this embodiment, the learning device 100 uses the measurement data of past communication quality measured at various locations (positions) to learn, by machine learning, a model that outputs statistical values (specifically, for example, the mean value and variance) of communication quality based on the position information indicating the position. The estimation device 200 uses the learned model learned by the learning device 100 to output the upper limit value and the lower limit value of the communication quality that fall within a desired distribution range from the position information.

[0017] Hereinafter, the configurations and operations of the learning device 100 and the estimation device 200 will be described in detail.

[0018] (Configuration of the Learning Device 100) FIG. 1 shows a configuration example of the learning device 100. As shown in FIG. 1, the learning device 100 includes an input unit 110, a communication quality DB (database) 120, a learning data calculation unit 130, a learning data DB 135, a learning unit 140, a model DB 150, and an output unit 160. Note that the communication quality DB 120 may be provided outside the learning device 100. The functional outlines of the respective units are as follows.

[0019] The input unit 110 inputs past history information of communication quality having position information and the measured value (measurement data) of communication quality at that position. The input history information is stored in the communication quality DB 120.

[0020] The learning data calculation unit 130 generates (calculates) learning data from the history information stored in the communication quality DB 120 and stores the generated learning data in the learning data DB 135.

[0021] The learning unit 140 trains a model using machine learning with the training data stored in the training data DB 135. The model in this embodiment is not limited to a specific type, but it is assumed that, for example, a neural network model will be used.

[0022] The model trained by the learning unit 140 is stored in the model database 150. Specifically, the model stored in the model database 150 consists of, for example, the function and weight parameters.

[0023] The models stored in the model DB150 are output from the output unit 160 and input to the estimation device 200.

[0024] As mentioned above, the learning device 100 and the estimation device 200 may be the same device. For example, the learning device 100 may include the functions of the estimation device 200. In this case, the learning device 100 can use a trained model to output the range of change in communication quality (upper limit, lower limit) in response to an inquiry (location information, etc.).

[0025] (Operation of the learning device 100) Next, the operation of the learning device 100 will be explained in detail with reference to the flowchart in Figure 2. Here, it is assumed that the communication quality DB 120 already stores historical information (measurement data) of communication quality measured at various locations.

[0026] <S101、S102> In S101, the learning data calculation unit 130 refers to the communication quality DB 120. In S102, the learning data calculation unit 130 obtains communication quality measurement data at a certain point p from the communication quality DB 120.

[0027] Measurement data at a certain point p refers to measurement data measured by the terminal while the terminal is at that point p. However, it is not limited to this; for example, the measurement data at a certain point p may also be measurement data measured by another device communicating with the terminal while the terminal is at that point p.

[0028] The measurement data acquired here includes, for example, location information of point p (e.g., latitude, longitude) and communication quality values ​​(e.g., received power, throughput, delay, jitter, packet loss). Note that the "communication quality values" may also be referred to as "communication quality".

[0029] For example, if the communication quality is defined as "received power, throughput, and latency," then statistical values ​​(mean, variance, etc.) will be calculated and learned for each of these, as explained below.

[0030] <s103> In S103, the learning data calculation unit 130 searches the communication quality DB 120 for and obtains measurement data for each point within X [m] of point p of the measurement data acquired in S102. The acquired data group is D p Let's assume that.

[0031] Note that X may be predetermined or input from the input unit 110. Each measurement data includes the communication quality at the measurement point.

[0032] <s104> In S104, the learning data calculation unit 130 is D p For each communication quality, the mean and variance are calculated. For the variance, either the sample variance or the unbiased variance may be calculated, or both may be calculated. The calculated mean and variance, along with the location information (p), are stored in the training data DB135.

[0033] Note that the mean and variance are examples of one or more statistical values. For example, if either the mean or variance is known (or if a specific value is used as an assumption), then only one of the mean or variance may be calculated.

[0034] <s105> The learning data calculation unit 130 repeatedly executes S102 to S104 while changing point p until processing of all target measurement data is completed (until S105 becomes Y).

[0035] <s106> The learning unit 140 uses the "location information, mean, and variance" stored in the learning data DB 135 to train a model for each communication quality. The trained model is stored in the model DB 150.

[0036] In this embodiment, we train two models: one that takes location information as input and outputs the mean value (which we will call the mean model), and another that takes location information as input and outputs the variance (which we will call the variance model).

[0037] For example, with an average value model, location information is input, and the parameters are adjusted to minimize the error between the model's output and the mean value of the correct answer. The same applies to other models.

[0038] Furthermore, as mentioned above, training a mean model and a variance model is just one example. Alternatively, a model that outputs the standard deviation could be trained instead of the variance model.

[0039] Alternatively, a model may be trained that outputs both the mean and variance from location information. Alternatively, a model may be trained that takes location information and a desired distribution range P[%] (described later) as input and outputs the upper and lower bounds for a normal distribution where the probability of a value falling within that range is P.

[0040] (Estimation device 200) Next, the estimation device 200 will be described. Figure 3 is a diagram of the configuration of the estimation device 200. As shown in Figure 3, the estimation device 200 comprises an input unit 210, an estimation unit 220, a model DB 230, and an output unit 240.

[0041] Model DB230 stores the trained models that have been trained by the learning device 100.

[0042] Figure 4 shows an example of the internal configuration of the estimation unit 220, assuming that the mean value model and the variance model are learned by the learning device 100.

[0043] As shown in Figure 4, the estimation unit 220 comprises a mean estimation unit 221, a variance estimation unit 222, and a calculation unit 223. The mean estimation unit 221 reads the mean model from the model DB 230 and estimates the mean from the location information using the mean model.

[0044] The variance estimation unit 222 reads a variance model from the model DB 230 and uses the variance model to estimate the variance from the location information.

[0045] The calculation unit 223 calculates the upper and lower bounds of a normal distribution with an estimated mean and variance, where the probability of a value falling within that distribution is P[%].

[0046] (Operation of estimation device 200) The operation of the estimation device 200, which has the configuration shown in Figures 3 and 4, will be explained with reference to the flowchart in Figure 5. Here, "communication quality" is assumed to be a specific type of communication quality (e.g., throughput). You may also input in S201 which type of communication quality range you want to know the range of change for.

[0047] <s201> In S201, the input unit 210 receives location information (e.g., latitude, longitude) for which the range of change in communication quality (specifically, upper and lower limits) is to be estimated, and a desired distribution range P[%]. This input information may be received, for example, as an inquiry from a terminal connected to the estimation device 200 via a network.

[0048] The location information is input to the mean estimation unit 221 and the variance estimation unit 222 of the estimation unit 220, respectively. The desired distribution range P[%] is input to the calculation unit 223 of the estimation unit 220. Alternatively, the distribution range P[%] may be pre-set in the calculation unit 223 as a predetermined value.

[0049] <s202> In S202, the mean estimation unit 221 uses an mean model to obtain a predicted mean value μ from the location information. The predicted mean value μ is input to the calculation unit 223.

[0050] <s203> In S203, the variance estimation unit 222 uses a variance model to predict the variance σ from the location information. 2 Obtain the predicted value of the variance σ. 2 This is input to the calculation unit 223.

[0051] <s204> In S204, the calculation unit 223 calculates the predicted value μ and the predicted value σ. 2 , and from the desired distribution range P[%], the μ and σ 2 Assuming a normal distribution, we calculate the upper and lower bounds of the distribution for which the probability of a value falling within that distribution is P[%]. The upper bound may be called the higher quality, and the lower bound may be called the lower quality. Figure 6 shows an illustration of P[%] and the upper and lower bounds. Note that Figure 6 shows a range centered on the mean, but this is just one example. Depending on the application, you may also calculate the upper and lower bounds for a range that is biased towards better (or worse) quality than the mean.

[0052] <s205> The calculated upper and lower limits are passed to the output unit 240 and output from the output unit 240. For example, the output unit 240 includes a display, and the upper and lower limits are displayed on the display. Alternatively, the output unit 240 may transmit the upper and lower limits to the requesting terminal via the network.

[0053] Alternatively, the estimation device 200 may not output the upper and lower limits, instead retaining them. The retained upper and lower limits can be read from an external device as needed.

[0054] (Example hardware configuration) Any of the devices described in this embodiment (learning device 100, estimation device 200) can be realized, for example, by having a computer run a program. This computer may be a physical computer or a virtual machine on the cloud.

[0055] In other words, the device can be realized by using hardware resources such as the CPU and memory built into a computer to execute a program corresponding to the processing performed by the device. The program can be recorded on a computer-readable recording medium (such as portable memory), saved, and distributed. It can also be provided via a network, such as the Internet or email.

[0056] Figure 7 shows an example of the hardware configuration of the computer described above. The computer in Figure 7 has a drive device 1000, an auxiliary storage device 1002, a memory device 1003, a CPU 1004, an interface device 1005, a display device 1006, an input device 1007, an output device 1008, etc., all of which are interconnected by a bus BS. The computer may also be equipped with a GPU.

[0057] The program that enables processing on the computer is provided, for example, on a recording medium 1001 such as a CD-ROM or memory card. When the recording medium 1001 containing the program is set in the drive device 1000, the program is installed from the recording medium 1001 to the auxiliary storage device 1002 via the drive device 1000. However, the program does not necessarily have to be installed from the recording medium 1001; it may also be downloaded from another computer via a network. The auxiliary storage device 1002 stores the installed program as well as necessary files and data.

[0058] The memory device 1003 reads and stores a program from the auxiliary storage device 1002 when a program startup command is received. The CPU 1004 implements the functions related to the memory device 1003 according to the program stored in the memory device 1003. The interface device 1005 is used as an interface for connecting to a network, etc. The display device 1006 displays a GUI (Graphical User Interface) etc. generated by a program. The input device 1007 consists of a keyboard and mouse, buttons, or a touch panel etc., and is used to input various operation commands. The output device 1008 outputs the calculation results.

[0059] (Effects of the embodiment) The technology described in this embodiment makes it possible to estimate the range of changes in wireless communication quality, and to make decisions on the use of wireless access considering the lower and upper limits of quality estimation based on location information and their probabilities. In other words, even in the case of wireless communication by a moving object whose location changes, it becomes possible to estimate with the greatest possible accuracy whether the desired communication quality can be continuously met.

[0060] The following additional information is disclosed regarding the embodiments described above.

[0061] <Note> (Additional note 1) An estimation device for estimating the range of changes in communication quality at a certain location, Memory and At least one processor connected to the memory, Includes, The aforementioned processor, Input location information indicating the aforementioned location, Using a trained model, one or more statistical values ​​related to the communication quality are estimated from the location information, and the range of change is estimated based on the one or more statistical values. Estimation device. (Additional note 2) The one or more statistical values ​​are the mean and the variance, and the processor uses the distribution having the mean and the variance to estimate the range of the change. The estimation device described in Appendix 1. (Additional note 3) The processor estimates upper and lower limits corresponding to a desired range in the distribution as the range of change. Estimation device as described in Appendix 2. (Additional note 4) A learning device for learning a model used to estimate the range of changes in communication quality at a certain location, Memory and At least one processor connected to the memory, Includes, The aforementioned processor, From the communication quality measurement data at each location, one or more statistical values ​​are calculated for each location. The model is trained by using the location information of each position and the one or more statistical values ​​mentioned above as training data. Learning device. (Additional note 5) An estimation method performed by an estimation device for estimating the range of changes in communication quality at a certain location, The steps include: inputting location information indicating the aforementioned location; The steps include: using a trained model to estimate one or more statistical values ​​related to the communication quality from the location information, and estimating the range of change based on the one or more statistical values; An estimation method comprising the following: (Additional note 6) A learning method performed by a learning device that learns a model used to estimate the range of changes in communication quality at a certain location, The steps include: calculating one or more statistical values ​​for each location from the communication quality measurement data at each location; The step of training the model by using the location information of each position and the one or more statistical values ​​mentioned above as training data. A learning method that includes [the following features]. (Additional note 7) A non-temporary storage medium storing a program for causing a computer to function as a component of the estimation device described in any one of the appendices 1 to 3. (Additional note 8) A non-temporary storage medium storing a program for causing a computer to function as a component in the learning device described in Appendix 4.

[0062] Although this embodiment has been described above, the present invention is not limited to this specific embodiment, and various modifications and changes are possible within the scope of the gist of the invention as described in the claims. [Explanation of Symbols]

[0063] 100 Learning Devices 110 Input Section 120 Communication Quality Database 130 Learning Data Calculation Unit 135 Training Data Database 140 Learning Department 150 Model DB 160 Output section 200 Estimation device 210 Input section 220 Estimation section 221 Mean Estimation Section 222 Variance estimation part 223 Calculation Department 230 Model DB 240 Output section 1000 drive unit 1001 Recording media 1002 Auxiliary storage 1003 Memory device 1004 CPU 1005 Interface device 1006 Display device 1007 Input device 1008 Output device

Claims

1. An estimation device for estimating the range of changes in communication quality at a certain location, An input unit for inputting position information indicating the aforementioned position, An estimation unit that uses a trained model to estimate the mean and variance as one or more statistical values ​​relating to the communication quality from the location information, and estimates the upper and lower limits of the range in which the probability of a value falling within a normal distribution having the mean and variance is a desired distribution range P [%] as the range of change. An estimation device equipped with the following features.

2. A learning device for learning a model used to estimate the range of changes in communication quality at a certain location, A learning data calculation unit that calculates the mean and variance as one or more statistical values ​​for each location from the communication quality measurement data at each location, The system includes a learning unit that learns the model by using the positional information of each location, the mean value, and the variance as training data. The aforementioned model takes location information and a desired distribution range P[%] as input and outputs upper and lower limits of the range in which the probability of a value falling within the distribution range P[%] is within a normal distribution having a mean and variance corresponding to the location information. Learning device.

3. An estimation method performed by an estimation device for estimating the range of changes in communication quality at a certain location, The steps include: inputting location information indicating the aforementioned location; The steps include: using a trained model, estimating the mean and variance as one or more statistical values ​​relating to the communication quality from the location information; and estimating the upper and lower limits of the range in which the probability of a value falling within a normal distribution having the mean and variance is a desired distribution range P [%] as the range of change; An estimation method comprising the following:

4. A learning method performed by a learning device that learns a model used to estimate the range of changes in communication quality at a certain location, The steps include: calculating the mean and variance as one or more statistical values ​​for each location from the communication quality measurement data at each location; The method comprises the step of training the model by using the positional information of each location, the mean value, and the variance as training data. The aforementioned model takes location information and a desired distribution range P[%] as input and outputs upper and lower limits of the range in which the probability of a value falling within the distribution range P[%] is within a normal distribution having a mean and variance corresponding to the location information. Learning methods.

5. A program for causing a computer to function as a component of the estimation device described in claim 1.

6. A program for causing a computer to function as a component of the learning device described in claim 2.

Images