CN111901893A - Method and apparatus for operating a wireless communication network - Google Patents

Abstract

A method for operating a wireless communication network (4) comprising a plurality of communication interfaces (iT1-iT3, iAPs) that are spatially separated from each other, wherein the method comprises: according to an estimated future connection quality ( q(t2)) configure at least a part of the communication interfaces in the plurality of communication interfaces (iT1-iT3, iAP). It also relates to an apparatus for operating a wireless communication network.

Description

Method and apparatus for operating a wireless communication network

technical field

The present invention relates to a method for operating a wireless communication network and an apparatus for operating a wireless communication network.

Background technique

Knowing the quality of the connection between the two points is either performed by measurements for the current moment.

Furthermore, it is known from radio network planning to estimate connection quality based on an environment model.

SUMMARY OF THE INVENTION

The problems of the prior art are solved by a method according to claim 1 and a device according to the dependent independent claim. Advantageous developments are specified in the dependent claims.

According to a first aspect of the present specification, a method for operating a wireless communication network comprising a plurality of communication interfaces that are spatially separated from each other, wherein the method comprises: determining which of the plurality of communication interfaces the actual position in space where the respective communication interfaces of the the future position in space where the respective communication interfaces of the respective communication interfaces are located at the future moment; at least two of the plurality of communication interfaces are respectively determined at the second moment according to the actual position, according to the actual connection quality and according to the future position future connection quality of radio connections established or achievable between communication interfaces; and configuring at least a portion of the plurality of communication interfaces according to the estimated future connection quality.

The future connection quality represents a metric that can simply be used to reconfigure the communication network. Hence, an advantageous mapping of actual state and expected future state to said future connection quality is made.

Using the estimated future connection quality, a targeted reconfiguration of the wireless communication network is advantageously performed to ensure uninterrupted communication while high connection quality. The prediction of the future connection quality makes it possible to increase the availability of the communication network as a whole, especially in scenarios where objects of the communication interface are moved.

Said estimated future connection quality advantageously complements and completes the future representation of space, such that the prediction foresees connection losses or degradation of said connections and can be prevented by corresponding countermeasures in the form of configurations. deterioration.

In the field of industrial automation, for example, new real-time applications can be realized or first made possible by the proposed method.

An advantageous example is characterized in that the method comprises propagating the determined actual position, the determined actual connection quality and the future position by a learned artificial neural network, wherein in the input area of the artificial neural network The actual location, the actual connection quality and the future location are provided as input variables in , and wherein a prediction that can be used to determine the estimated future connection quality is provided in an output region of the artificial neural network.

Advantageously, the artificial neural network is used to take into account the scene dynamics contained in the mapped actual situation when reconfiguring the communication network. The provision of the prediction allows an estimation of the future connection quality.

An advantageous example is characterized in that the method comprises determining the actual position and/or the future position from sensor data provided by sensors that at least partially observe the space.

Advantageously, for example, the information representing the actual position of the respective communication interface is enriched by the sensor data or is only provided by the sensor data. Therefore, information is added by adding the sensor data.

An advantageous example is characterized in that the method comprises: identifying from the sensor data at least one object and its position in the space; linking the identified object with one of the plurality of communication interfaces ; and determining the position of the one communication interface according to the position of the object linked with the one communication interface of the plurality of communication interfaces.

The position determination for the communication interface is advantageously improved by the proposed data fusion.

An advantageous example is characterized in that the method comprises: based on a comparison of physical radio channels already used for wireless communication by means of the plurality of communication interfaces and physical radios available for wireless communication by means of the plurality of communication interfaces Monitoring of the channel, determining the actual location and/or the actual connection quality and/or the future location.

The actual position and/or the estimated future position of the communication interface is determined by the provided monitoring without additional hardware being provided for this purpose. Instead, existing hardware can be used.

An advantageous example is characterized in that configuring at least a part of said communication interfaces comprises establishing a first radio connection between at least two communication interfaces of said communication network; according to said estimated future connection quality establishing a second radio connection between the at least two communication interfaces; and switching a communication path from the first radio connection to the second radio connection.

Advantageously, this type of preventive radio channel switching can be used to prevent an undesired interruption of the first radio connection or a degradation of the connection quality of the first radio connection. The availability of communication within the wireless communication network is thereby increased.

A second aspect of the present specification relates to an apparatus for operating a wireless communication network, wherein the apparatus comprises at least one processor and a memory having computer program code, wherein the computer program code is executed with the processor such that all The device determines the actual position in space where the respective communication interface of the plurality of communication interfaces is located; determines the actual connection quality of the radio connection established or can be established between the two communication interfaces of the plurality of communication interfaces respectively; determines It is expected that a future position in space where each of the plurality of communication interfaces is located at a future moment; according to the actual position, according to the actual connection quality and according to the future position future connection quality of radio connections established or achievable between at least two of the plurality of communication interfaces; and configuring at least a portion of the plurality of communication interfaces according to the estimated future connection quality.

Description of drawings

In the attached image:

Figure 1 shows a schematic diagram of an exemplary scene with spaces;

Figure 2 shows the training of a schematic representation of an artificial neural network;

Figure 3 shows the use of a schematic representation of a trained artificial neural network; and

Figures 4 and 5 show representations of the space, respectively.

Detailed ways

FIG. 1 shows a schematic diagram of an exemplary scenario 2 with a space in which a wireless communication network 4 operates. The wireless communication network 4 comprises eg a number of mobile terminals T1, T3 and/or fixed terminals such as terminal T2. The communication network 4 also includes access points AP.

The terminals T1 to T3 and the access point AP respectively comprise at least one communication interface iT1 to iT3 and iAP. The communication interfaces iT1 to iT3 are, for example, designed to establish a radio connection to each other, for example a radio connection C23 . The communication interface iAP provides access to another network area NA. Thus, for example, a radio connection C1AP is established between the communication interface iAP and the communication interface iD1. The access point AP thus provides the terminals T1 to T3 with the possibility to communicate via the communication path CP with the remotely arranged network elements NE. In this case, the communication path CP includes the radio connection C1AP and at least one other connection Cx between the access point AP and the network element NE arranged remotely. The network elements NE are, for example, not part of the communication network 4, but are arranged in another network area NA of the respective enterprise to which the communication network 4 is also assigned, or in the wide area network.

In addition to the terminals T1 to T3 representing radio objects and at least one access point AP, the scenario 2 also includes other passive objects O1 and O2 which affect the radio traffic between the terminals T1 to T3 and the access point AP . In the example shown, the radio connection C1AP is not covered by one of the passive objects O1 and O2. The radio connection C1AP is, for example, less affected by signal attenuation than the radio connection C3AP between the communication interfaces iT3 and iAP, since the communication interfaces iT1 and iAP have line-of-sight contact, ie there are no passive objects between the communication module iT1 and the iAP. Or an imaginary straight-line connection directly covered by a radio object. In the case of the radio connection C3AP, a higher signal attenuation must be expected compared to the radio connection C1AB.

If the mobile terminal T1 moves along the trajectory tT1 with the assigned communication module iT1 from the first moment, the mobile terminal T1 reaches the position pT1 at the second moment in the future and is blocked by the passive object O1 from the point of view of the access point AP occlude. That is to say, the communication modules iT1 and iAP have no line-of-sight contact at the second moment. Therefore, an increase in the signal attenuation for the radio connection C1AP must be assumed at the second point in time. In another example, it is also conceivable that one of these mobile terminals is located between the first terminal T1 and the access point AP at the second time instant.

The scene 2 or the allocated space is at least partially detected by at least one sensor S, the sensor data SD of which are fed to the device 100 . The sensor S is, for example, a camera, a radar sensor, a lidar sensor or an ultrasonic sensor. Of course, multiple different types of sensors can also be used to monitor the scene. Device 100 includes at least one processor and memory on which a computer program is stored. When the computer program is executed on the processor, the method steps explained in this specification are performed.

The access point AP or its communication interface iAP allows the observation of used and unused radio channels or the entire radio traffic in the communication network 4 . The radio observations Obs are transmitted to the device 100 and used to determine the first or second representative.

According to block 102 of the device 100, the actual spatial position p(t1) at which the respective communication interface of the plurality of communication interfaces iT1 to iT3, iAP is located at the first moment (especially the current moment) is determined. The actual position p(t1) is determined, for example, from the radio observations Obs and/or from the sensor signal SD.

Block 104 determines the actual connection quality q(t1) of the radio connections established between at least two of the plurality of communication interfaces iT1 to iT3, iAP, respectively, at said first time instant. These actual connection qualities q(t1) are determined from the radio observations Obs.

Block 106 determines the respective future spatial positions p(t2) at which the plurality of communication interfaces iT1 to iT3, iAPs are expected to be located at a second time instant later than said first time instant. This is done, for example, by observing the trajectory of the respective communication interfaces iT1 to iT3, iAP, wherein, for example, in addition to the actual position at the first point in time, the associated speed and direction can also be determined. Correspondingly, an estimated area is obtained, in which the respective communication interfaces of the communication interfaces iT1 to iT3, iAP may be located at the second instant in time. A position in the sense of the spatial position p(t2) can then be selected from the estimated region. Alternatively, the estimated area can also be determined by means of block 106 , possibly still with respective position-dependent probabilities. Alternatively or additionally, the communication interfaces iT1 to iT3, iAP can transmit the planned trajectory to the device 100, from which the respective future position is determined. In one example, the future position p(t2) is estimated from the actual position p(t1).

Block 108 determines the future connection quality q(t2) of the radio connections established or possible to establish between at least two of the plurality of communication interfaces iT1 to iT3, iAP, respectively, at said second time instant. The connection quality q(t2) is determined from the actual position p(t1), from the actual connection quality q(t1) and from the future position p(t2).

Block 108 includes, for example, an artificial neural network determining future connection quality q(t2).

Block 110 determines a configuration Conf of at least a portion of the plurality of communication interfaces iT1 to iT3, iAPs, based on the estimated future connection quality q(t2). Advantageously, it is thus possible, for example, to clear existing radio connections and to establish new ones before connection losses occur, which can be predicted from the future connection quality q(t2). At least some of the plurality of communication interfaces iT1 to iT3, iAP of the communication network 4 are configured accordingly.

Then, when a degradation of one of the possible radio channels at the second instant t2 is expected, ie, for example, loss of line-of-sight contact, the configuration Conf is determined and executed. Thus, for example, the communication takes place via a radio channel to be newly established which has a lower expected signal attenuation at the second instant than the existing radio connection, if the existing radio channel is in the If the second moment still exists. For example, the moving terminal switches to a radio connection at an access point, not shown, which is not the access point AP, even before entering said second time instant. The configuration Conf is distributed to the terminals T1 to T3 via the access point AP, for example. A partial or complete reconfiguration of the communication interfaces iT1 to iT3 and iAP is then performed before or upon entering said second time instant. This reconfiguration includes, for example, separation and establishment of radio channels, increase or decrease of reception strength and/or transmission strength and/or data rate and/or packet rate, switching of frequency bands, switching of modulation and coding schemes, and the like. Within the scope of the configuration, for example, coordination between existing access points takes place.

The apparatus for training the artificial neural network is shown in FIG. 3 . The training data E _train are provided in the form of recorded actual positions p(t1), actual connection qualities q(t1) and future positions p(t2).

The apparatus includes an artificial neural network 200 having an input layer. For time step i, pass the input tensor e ⁱ _train to the input layer. For the input E, the output O is determined in the form of a prediction of the output O or is known in advance. From the output O, a tensor o ⁱ _train is determined at time step i with observations that are assigned to the observations of the tensor e ⁱ _train . Each time series of input E is assigned to one of the three input nodes. In the forward path of the artificial neural network 200, at least one hidden layer follows the input layer. In this example, the number of nodes of the at least one hidden layer is greater than the number of input nodes. This number is considered a hyperparameter and is preferably determined individually. In this example, four nodes are provided in the hidden layer. The artificial neural network 200 is taught, for example, by a method according to gradient descent in the form of backpropagation. Therefore, the artificial neural network 200 is trained in a monitored manner.

Of course, the neural network 200 can also be taught in different ways. Thus, for example, the highest possible connection quality can be set as a learning target, and the artificial neural network 200 can be taught through monitored learning. Instead, in the case of Reinforcement Learning, use an objective function such as the lowest connection failure rate.

In the forward path, an output layer 202 is provided after the at least one hidden layer in this example. The predicted value is output at the output layer 202 . In this example, each predictor is assigned an output node.

At each time step i, a tensor o' ⁱ _train is determined in which the predicted values for that time step i are contained. In this example, the tensor is fed to the training device 204 along with a column vector o ⁱ _train of observations. The training device 204 is configured in this example to determine the prediction error by means of the loss function LOSS, in particular by means of the mean squared error, and to use this prediction error and to train the model by means of the optimizer, in particular the Adam optimizer. The loss function LOSS is determined in this example from the deviation, in particular the mean squared error, between the tensor o ^{i train of the observed values and the tensor o' i} _train ^of _the predicted values.

The training is interrupted as soon as the set criteria are reached. In this example, the training is interrupted when the loss no longer decreases over a number of time steps, ie in particular the mean squared error does not decrease.

和协方差

来评估所述模型。The test data is then fed into the model trained in this way. The model is generated by training using the training data. Use the test data to evaluate the model, especially with respect to the mean

and covariance

to evaluate the model.

According to the apparatus shown in Figure 3, a trained model is used to provide a prediction of connection quality for the input E delivered. For this purpose, the same data preprocessing steps are performed as in the case of the training data. For example, scaling and determination of input data and output data is performed. In this example, the determination is made during operation of the device 100 , ie in the operation of the wireless communication network 4 .

Digital images that may contain objects to be classified are input into the trained artificial neural network 200 . Based on this, the predicted value is determined. Based on this, the connection quality score is determined.

A means for determining the quality of the future connection is schematically shown in FIG. 3 . As described for training, for time step ⁱ , the column vector ei of input E is passed to the input layer by means of block 201 . After this, unlike training, prediction is performed by the apparatus 300 on the future connection quality according to the predicted value ^y'i .

，则所述连接质量指示未来视线连接，否则就不指示未来视线连接。此外，如果所述连接质量得分低于另一个阈值，则所述连接质量得分可能指示无法连接。阈值

优选是参数。该参数例如借助于对例如精度、召回率（Recall）的标准进行最大化来加以确定。例如，使用曲线下面积（Area under the Curve）AUC或接收器操作特性（Receiver Operating Characteristic）ROC标准。The connection quality score is derived for the future connection quality q(t2) and passed to the device 110 . In this example, if the connection quality score exceeds the threshold

, the connection quality indicates the future line-of-sight connection, otherwise it does not indicate the future line-of-sight connection. Furthermore, if the connection quality score is below another threshold, the connection quality score may indicate an inability to connect. threshold

Preferably parameters. This parameter is determined, for example, by maximizing criteria such as precision, recall. For example, use the Area under the Curve AUC or Receiver Operating Characteristic ROC criteria.

In particular, the instructions of the computer program implementing the artificial neural network 200 are provided for carrying out the described methods. It is also possible to set up dedicated hardware in which the trained model is mapped.

Figure 4 shows a representative R1 of the space monitored by sensors Sa and Sb. In the first representation, the vehicle V1 is moving in the direction r. There are respective line-of-sight radio connections 402, 404 between the two access points APa and APb and the industrial robot IR, which are not covered by objects O1, O2 or O3 located in the space.

A second representation R2 of the space is shown in FIG. 5 , which is in the future based on the first representation R1 and estimated by means of the neural network 200 of FIGS. 2 and 3 . The artificial neural network 200 thus determines that, at a future instant, ie at the second instant, the vehicle V1 breaks the line-of-sight connection 404 of FIG. 4 because it blocks the industrial robot IR with respect to the access point APa. A corresponding reconfiguration may, for example, include increasing the transmit power and the receive power of the respective communication interfaces of the industrial robot IR and the access point APb.

Claims (7)

1. A method for operating a wireless communication network (4) comprising a plurality of communication interfaces (iT 1-iT3, iAP) spatially separated from each other, wherein the method comprises:

determining a spatial actual position (p (t 1)) at which a respective communication interface of the plurality of communication interfaces (iT 1-iT3, iAP) is located,

determining an actual connection quality (q (t 1)) of a radio connection established or able to be established between two of the plurality of communication interfaces (iT 1-iT3, iAP), respectively,

determining a spatial future position (p (t 2)) at which a respective communication interface of the plurality of communication interfaces (iT 1-iT3, iAP) is expected to be located at a future time instant,

determining a future connection quality (q (t 2)) of a radio connection respectively established or able to be established between at least two of the plurality of communication interfaces (iT 1-iT3, iAP) at the second point in time from the actual position (p (t 1)), from the actual connection quality (q (t 1)) and from the future position (p (t 2)), and

configuring at least a part of the plurality of communication interfaces (iT 1-iT3, iAP) according to the estimated future connection quality (q (t 2)).

2. The method of claim 1, wherein the method comprises:

propagating the determined actual position (p (t 1)), the determined actual connection quality (q (t 1)) and the future position (p (t 2)) through a learned artificial neural network (200), wherein the actual position (p (t 1)), the actual connection quality (q (t 1)) and the future position (p (t 2)) are provided as input variables (E) in an input area of the artificial neural network (200), and wherein a prediction is provided in an output area of the artificial neural network (200) that can be used to determine the estimated future connection quality (q (t 2)).

3. The method according to any one of the preceding claims, wherein the method comprises:

determining the actual position (p (t 1)) and/or the future position (p (t 2)) from Sensor Data (SD) provided by sensors (S) at least partially observing the space.

4. The method of claim 3, wherein the method comprises:

identifying at least one object and its position in the space from the Sensor Data (SD),

linking the identified object with one of the plurality of communication interfaces (iT 1-iT3, iAP), and

determining a position (p (t1), p (t 2)) of said one of said plurality of communication interfaces (iT 1-iT3, iAP) from a position of an object linked to said one of said plurality of communication interfaces (iT 1-iT3, iAP).

5. The method according to any one of the preceding claims, wherein the method comprises:

the actual position (p (t 1)) and/or the actual connection quality (q (t 1)) and/or the future position (p (t 2)) are/is determined from monitoring of radio channels already used for wireless communication by means of the plurality of communication interfaces (iT 1-iT3, iAP) and radio channels that can be wirelessly communicated by means of the plurality of communication interfaces (iT 1-iT3, iAP).

6. The method of any preceding claim, wherein the configuring of at least a portion of the communication interfaces comprises:

establishing a first radio connection between at least two communication interfaces (iT 1-iT3, iAP) of the communication network (4),

establishing a second radio connection between the at least two communication interfaces (iT 1-iT3, iAP) according to the estimated future connection quality (q (t 2)), and

switching a Communication Path (CP) from the first radio connection to the second radio connection.

7. An apparatus (100) for operating a wireless communication network (4), wherein the apparatus (100) comprises at least one processor and a memory with computer program code, wherein execution of the computer program code with the processor causes the apparatus (100)

Determining a spatial actual position (p (t 1)) at which a respective communication interface of the plurality of communication interfaces (iT 1-iT3, iAP) is located,

determining an actual connection quality (q (t 1)) of a radio connection established or able to be established between two of the plurality of communication interfaces (iT 1-iT3, iAP), respectively,

determining a spatial future position (p (t 2)) at which a respective communication interface of the plurality of communication interfaces (iT 1-iT3, iAP) is expected to be located at a future time instant,

determining a future connection quality (q (t 2)) of a radio connection respectively established or able to be established between at least two of the plurality of communication interfaces (iT 1-iT3, iAP) at the second point in time from the actual position (p (t 1)), from the actual connection quality (q (t 1)) and from the future position (p (t 2)), and

configuring at least a part of the plurality of communication interfaces (iT 1-iT3, iAP) according to the estimated future connection quality (q (t 2)).

Images