WO2020099008A1

WO2020099008A1 - Method for generating a coverage map for a wireless communication system, and coverage map

Info

Publication number: WO2020099008A1
Application number: PCT/EP2019/075997
Authority: WO
Inventors: Holger Kiehne; Jens Schwardmann; Radu Circa; Frank Hofmann
Original assignee: Robert Bosch Gmbh
Priority date: 2018-11-15
Filing date: 2019-09-26
Publication date: 2020-05-22
Also published as: DE102018219585A1

Abstract

The aim of the invention is to provide a method for generating a coverage map (10) for a wireless communication system. The method is used to generate a coverage map (10) on the basis of passive parameters, said coverage map being advantageously usable for different wireless technologies and applications and additionally being advantageous with respect to energy consumption and data usage. This is achieved in that the coverage map (10) comprises a plurality of network quality indicators, wherein a prediction model (11) is generated in order to predict the network quality, and each of the network quality indicators is assigned a measurement point within the geographical region detected by the coverage map (10). At least one passive parameter is measured in order to calculate a network quality indicator at a measurement point, and the network quality indicator for the measurement point is calculated using the prediction model (11) and the at least one passive parameter.

Description

description

title

Method of creating a coverage map for a wireless

Com m u n i kati o nss yste m and coverage map

The invention relates to a method for producing a cover card for a wireless communication system and to a cover card produced using this method.

STATE OF THE ART

A number of methods are known from the prior art with which cover cards of wireless communication systems are created which provide information about the network quality.

From US 906 0318 B2, for example, a method is known in which a map is created which contains directly measured parameters such as signal strengths,

Contains data rates or packet loss during transmission.

Furthermore, from US 2011 130 135 A1 a method for detecting coverage gaps in a wireless communication network is known, which extracts relevant data from participants in the network and creates a coverage map therefrom.

A disadvantage of previously known methods is that to create a

Coverage map either active measurements have to be made, for example to measure a data rate, or parameters from passive measurements, such as the signal strength, are used and these are entered directly into the coverage map.

An active measurement is to be understood as a measurement in which a transfer of test data is necessary, for example to measure a data rate. In networks with limited data volume such as The transmission of additional test data can therefore be expensive for mobile radio networks. In addition, active measurements put a strain on the network without any user data being transmitted and therefore also result in a higher energy requirement.

In the case of a passive measurement, local measured values are read out, such as a signal strength. No data transmission is necessary for the measurement of passive parameters. The local measured values can

can be read directly from a network modem, for example. However, signal strength information cannot be used directly by every application because it depends on the underlying radio technology, which is usually unknown to an application.

DISCLOSURE OF THE INVENTION

The object of the present invention is to create and provide an estimated prediction of network indicators and data such as data rates with the aid of a coverage map based on passive parameters, which can be evaluated for various radio technologies and applications and is also advantageous in Regarding energy and data consumption.

In particular, the prediction for a known route can be a

Vehicle. This object is achieved by the method for creating a cover card for a wireless communication system according to claim 1 and by the method created by the method

Cover card according to claim 10.

Appropriate embodiments of the invention are specified in the subclaims.

In order to enable network quality indicators to be estimated, a predictive model is required that is based on actively and passively measured data. Once the model has been created, only passive measured values are required to use the model. In contrast to the prior art, the coverage map is created on the basis of passively measured parameters such as signal strengths, which are used with the aid of a prediction model in order to estimate network quality indicators for forming the coverage map. Active measurements of are preferred only for the initial formation of the prediction model

Netw orkq u al ities i n d i cato re n necessary.

Network quality indicators are parameters that describe the network quality for an application. Examples of this are data rate, latency, jitter or packet loss rate.

According to the invention, a method for creating a coverage map for a wireless communication system is provided, the coverage map comprising at least one network quality indicator. A prediction model for predicting the network quality is created, the network quality indicators each being assigned to a measuring point within the geographical area covered by the coverage map. To calculate a

At least one passive parameter is measured at a measurement point at the network quality indicator, the network quality indicator being calculated for the measurement point by means of the prediction model and the at least one passive parameter.

With the help of the prediction, for example, data transmissions can be prioritized, so that in areas with poor coverage only important ones

Data transfers take place and those whose requirements regarding network quality are not met are not even started. The invention can thus contribute to higher-priority applications functioning better and additional data volumes being used more efficiently by avoiding unnecessary data transmissions by applications whose requirements are not met. In addition, services that require a certain data rate can store a larger amount of data before driving through areas with poor coverage, in order to achieve a higher quality of service. By matching the requirements of applications that transmit data with the predicted network quality, individual network interfaces in a client device can also be deactivated To save energy if these interfaces are not required in the foreseeable future to meet the requirements of all applications or if it is foreseeable that the interfaces will not allow any data transmission in the foreseeable future.

A first aspect of the invention relates to the estimation of

Network w itk a u itity i d i cato based on passive parameters that are stored in a coverage map.

A second aspect of the invention relates to predicting network quality with a coverage map server based on the coverage map of estimated network quality indicators. The network quality is preferably predicted along a known route or along at least one

Most Probable Path.

The route is preferably determined with the aid of a navigation device.

The wireless communication system is preferably a mobile radio network.

To create the prediction model, at least one active parameter, which is in particular a data rate, is preferably measured for each measuring point. A passive parameter is preferably a signal strength parameter.

The measurement of the at least one passive parameter and the calculation of the network quality indicator are preferably repeated cyclically or initiated by an inquiry. When measuring a passive parameter, the associated position is determined so that the passive parameter and the associated position form the measurement point. The measurements can be repeated periodically, for example once per second.

Passive parameters can preferably be measured at any time (e.g. also by the same vehicles that make inquiries). The calculation of the

Mains quality indicator can preferably also be triggered when new passive measured values are added. For each measuring point, the passive parameters and the determined network quality indicator are preferably stored in a memory. The coverage map formed in this way can preferably be stored on a coverage map server, which aggregates and processes measurement data collected by measuring devices in order to estimate network quality indicators and to store them in the map. The measuring devices can preferably be from a mobile

Communication means, which may have a localization device. Such a communication system can be, for example, a TCU (Telematics Control Unit). With the aid of the coverage map, a prediction of the network quality for a previously known route can preferably be created, which in turn can be used in a CCU.

A description is provided below for easier understanding

Prediction of data given. However, the description can be applied to other network quality indicators such as latency, jitter or packet loss rate.

The measurement data are preferably collected by measurement devices which contain at least one network device (e.g. mobile radio modem), one device for localization (e.g. via GPS), a processor and a cache. The individual types of measurement data can preferably be transmitted separately and are then combined in the coverage card server in measurement points. Alternatively, they can preferably be brought together directly by the measuring device.

A measurement point is a position to which one or more passively measured parameter values such as signal strengths are assigned. Additionally, actively measured parameters such as data rates can preferably be included.

The signal strengths are read from the network device and the position from a localization device. The network device is preferred,

for example a cellular modem, and the localization device, for example a GPS receiver, arranged in one module. The module can preferably be installed in a computer. For active measurement of

Network quality indicators such as data rates or latencies can preferably be used with software such as iperf and ping. The type of signal strength parameters depends on the underlying one

Radio technology. For mobile telephony with LTE (Long Term Evolution), the parameters SINR (Signal-to-Interference-plus-Noise Ratio), RSSI (Received Signal Strength Indicator), RSRP (Reference Signal Received Power), CQI (Channel Quality Indicator) or RSRQ (Reference Signal Received Quality) can be used. For mobile radio with UMTS (Universal Mobile

Telecommunications System), the parameters RSCP (Received Signal Code Power), RSSI, EC / NO (RS C P / total received power) or CQI can preferably be used. For mobile radio with GSM (Global System for Mobile

Communications), the parameters RSSI, C / l (carrier interference ratio) and BER (bit error rate) can preferably be used.

In addition to signal strengths, further parameters from

Network modem, such as the channel bandwidth, the ID of the cell, the registration status or information from the

Control channel. For WLAN (Wireless Local Area Network), RSSI or the channel bandwidth can preferably be used. The examples mentioned

Parameters can be collected in different combinations. The active measurements can preferably be carried out initially by a measurement fleet.

Measured data rates and the associated passively measured values are used to form the prediction model. In addition, further parameters can preferably be used, which are derived, for example, from the

Time or place of the measurement. Examples of this are given below.

The measurement points are preferably arranged in a group based on the passive parameters assigned to them. By doing

If data rate measurements are grouped based on their associated passive values, individual CCDFs can be formed. In order to be able to group the data rates, preferably not only one but also two or more signal strength parameters can be used. Certain parameters require that multiple coverage cards be made. These include, in particular, the provider of network access (e.g. cell phone provider or provider of a WLAN hotspot) and radio technology (eg LTE, UMTS, WLAN). Therefore, the collected data is preferably grouped first based on these parameter combinations, in order to then produce separate coverage cards. So preferably one is created

Cover card for Vodafone and LTE, one for Vodafone and UMTS, one for Telekom and LTE, etc.

The measurements grouped in this way are called data groups below. In order to then use the prediction model and to create a coverage map for each data group, the geographical area covered by the coverage map is preferably divided into sub-areas, each

Measuring point is assigned to a sub-area. For example, partial areas can have a square shape. To calculate the prediction of a

Sub-area is assigned for each measuring point assigned to the sub-area of the current data group from the measuring points in a sub-area

Network quality indicators use the prediction model to calculate a prediction, with the median preferably being calculated from the measurement points in a sub-area associated with network quality indicators. A prediction value is then calculated for each sub-area as well as optionally a confidence value that describes how likely the prediction is exceeded.

With the model described above, this is done by selecting the grouping and the associated CCDF based on the passive measurements of the respective point. A specific quantile is then read from this as a prediction for this point. The median is then determined from all predictions of a sub-area, which is used as an estimate of the data rate for the sub-area. This prediction is then preferably stored for the sub-area as a prediction for the respective data group, for example in a database.

To influence how pessimistic the prediction turns out, the quantile used to read a data rate from the CCDF of the model can be preferred is used, be chosen accordingly large or small. The value of the quantile is also preferably used as an initial value for the

Confidence used in prediction.

The network quality indicators for waypoints in a known route are preferably called up from the memory by means of the cover map. A prediction of data rates for a route can preferably be requested from the coverage map server. Such an inquiry preferably consists of individual waypoints (positions) of the route and additional ones

Information such as the mobile phone provider, radio technology or other relevant parameters.

Alternatively, a prediction model is preferably created per radio technology for all providers. This can be adapted to different providers, in particular by taking into account the maximum achievable data rate of a provider and using this data rate as the maximum prediction of the provider. This approach does not use data rates from providers that limit data rates when building models.

When the measurements are grouped, further groups are preferably generated, depending on technology-dependent parameters such as the channel bandwidth, the frequency band or different versions of a radio standard.

When creating a prediction, the capabilities of the communication means and further properties of the connection for which the query is being created are preferably taken into account. These include in particular

Channel bandwidth of the connection, the ability of the communication medium and remote station to carry out carrier aggregation as well as other technology-dependent parameters.

An alternative possible prediction model preferably uses a CDF (Cumulative Distribution Function) of data rates per passive measured value.

For example, if a passive parameter has a value range from 1 to 10 and only whole numbers occur, 10 different CDFs formed by using the associated data rates for the formation of a CDF for each of the values. In addition to integers, the

Value range also have non-integer values. In order to obtain a prediction for a partial area, a weighted group, preferably the corresponding CDF, is used for each of the measured values occurring there and added. The weighting is based on the relative frequency of occurrence of the measured value. This results in a total CD F per sub-area, which can preferably be used with a quantile in order to obtain a data rate for the sub-area.

In addition, the following alternatives and extensions can preferably be used:

Preferably, data prediction can be based on passive

A different model can be used, e.g. is based on a neural network or other machine learning method.

The recorded measurements are preferably temporarily stored on the measuring device and not all data are transmitted. Around

For example, to protect the driver's privacy, data can be transmitted at random times or only at certain locations. Since only parts of a trip are transmitted to the server, a reconstruction of the route traveled is prevented.

Another preferred measure for anonymizing the measurement data involves removing the exact time stamp, so that e.g. only the hour in which a measurement was taken is transferred to the server.

By preferably informing the coverage map server of areas to the measuring devices in which there are not yet enough measurements available, unnecessary measurements can be prevented. This message preferably contains a limitation of the area as well as optional information on which measurement data is still required there (e.g. which provider or which radio technology). Additionally or alternatively, one or more of the following influences can preferably be used to improve the prediction:

Influences due to the weather can be taken into account by improving or worsening the forecast and possibly the confidence value for a sub-area based on the current or expected weather conditions. The values are improved in clear weather, and deteriorated in heavy rain, fog or snow.

The weather data can be obtained from an online service.

Alternatively or in addition, the weather data can also be collected locally by the vehicle. For example, information can be collected about whether the windshield wiper is switched on, for example about the speed level of the windshield wiper, based on a rain sensor, based on a camera image or based on the friction of the current driving surface.

As an alternative to influencing the forecast by weather data, different maps can be used to show different ones

Weather conditions to be taken into account.

Time dependency: Depending on the day of the week or the time, either different coverage cards can be created or the value of the forecast can be influenced. E.g. For example, the forecast value can deteriorate at times when increased traffic is expected (e.g. during rush hour traffic) or improve at night.

Traffic volume: Based on known or expected traffic

Traffic in a sub-area improves the forecast (for below-average traffic) or worsens (for above-average traffic). The traffic volume can be via an online service, via local historical data or via a

Estimation based on vehicle sensors. Travel times for public transport: An online service can be used to query times when, for example, trains or buses pass through a certain area and thus an increased load in the wireless communication system can be expected. If a forecast is requested for such a time, the value stored in the map for the

Forecast worsened.

In addition to explicit measurements, active measurements can also be carried out

User data transfers are used for additional

Get data rate measurements. If the observed data rate lies above the prediction of the current sub-area, the measurement can be included in the prediction. If the measurement is below the prediction, a look-up table can be used with services whose data rate is not artificially limited in order to rule out that the measured data rate was throttled by a limitation of the service itself. If such a table is used, data rates lower than the prediction can also be used to be included in the prediction. On the basis of the deviation of the measured data rates from the predicted data rate, the confidence value, if present, can also be reduced or increased.

It is reduced if the proportion of measurements that are larger than the prediction is smaller than the confidence value. It is increased if the proportion of measurements that are larger than the prediction is larger than the confidence value.

Variable sub-area size: A variable can be used to cover smaller fluctuations and small areas with poor coverage

Sub-area size can be used. A size is used as the starting value (e.g. 50 m width and length per sub-area), but a sub-area can be divided into several smaller sub-areas. In addition, neighboring sub-areas of one size can become a larger sub-area

be summarized. The following conditions can be used to decide whether a sub-area should be subdivided or several sub-areas should be combined into a larger sub-area: o The variance and the mean value of the predictions of a sub-area or the passive measured values of a sub-area: If the variance is large, a sub-area is divided. If the variances of at least two neighboring sub-areas of the same size are small and the mean values are similar, they become

summarized.

o The number of measured values or predictions of a sub-area: If there is a high number of passive measured values or predictions in a sub-area, the sub-area is divided into several smaller sub-areas, provided that after the division there is a minimum number of measured values or predictions for each of the smaller sub-areas would.

Minimum number of measured values per subarea: A forecast is only carried out for those subareas in which there is a minimum number of valid passive measured values (e.g. only subareas in which there are at least five valid passive measured values).

Up-to-dateness of the measured values: Only measured values are used for a forecast that have a certain actuality (e.g. not older than 30 days). The number of days can also be reduced for sub-areas in which there are many measured values.

Certain positions with particularly good or bad reception:

Based on map data, locations can be identified where it is known that the geographical dependency of the network quality indicators is particularly high there. Examples of this are tunnel entrances or exits. A smaller sub-area size can be used there or the confidence value reduced if necessary.

Differentiation of different areas based on map data: Map data contain classifications of different areas such as urban areas, rural areas or forest areas that can be used to determine the

Adjust prediction and confidence level. In urban areas, the geographical dependence of network quality indicators is greater than in rural areas with little or no development, so that the level of confidence in urban areas is reduced and increased in rural areas. In addition, different models can be used for the different areas. A model for urban areas is only formed from data that was also collected in the urban area. A model is created for each area that is distinguished.

Number of sub-areas per cell ID: The number of

Subareas where a cell ID has been observed can be used to estimate how much the cell is under stress. If a cell ID has been observed in a very large number of areas, it is likely that the cell is often under heavy load. In this way, the prediction calculated from the model and, if appropriate, the confidence value can be reduced for sub-areas in which this Cell ID occurs. Alternatively, different models can be set up, depending on the number of sub-areas in which a Cell ID occurs. For example, a model can be created for cell IDs that occur in less than 20 sub-areas, and a model for cell IDs that occur in 20 sub-areas or more.

GPS accuracy: In the case of GPS positions, values are specified in addition to the position itself, which indicate the estimated accuracy of the position. This accuracy can be used as an influencing parameter for the prediction, since if the accuracy is low, there are likely to be many interfering influences that can also interfere with mobile radio reception. Therefore, the prediction and the confidence can deteriorate for areas where the GPS accuracy was often very low.

Direction of travel: The GPS position also contains an orientation (English "Heading"), which indicates the direction of movement. When calculating the prediction, these values can also be used to e.g. for a

Sub-area that contains a road with lanes in opposite directions to make separate predictions if these differ noticeably from one another. If this extension is used, a query for a prediction would also have to contain this parameter or it would have to come from the

Order of the positions of the request can be determined. By working with a network provider, the actual number of users in a cell can also be used. When creating active measurements, these can be stored as parameters. Then, when answering a prediction request, the number of currently active users can be used to worsen the prediction, provided that more users are active in the cell than on average at the time of the active measurements.

Number of storeys of buildings in a sub-area: In addition to the differentiation by area, the type of development can be differentiated if this data is available. So the prediction for

Subareas containing buildings with a high number of floors deteriorate, assuming that there are more reflections there and thus the network quality indicators are worse than in areas with lower buildings. Alternatively, if different models are used for different areas, the urban area can be subdivided again into areas with predominantly low buildings and areas with predominantly tall buildings. Thus, two separate models could be created for these two types of development. In addition to the

Reception conditions can indirectly influence the network load of buildings. It can be expected that there will be more users in train stations or buildings with many floors.

Season: In forest areas, the reception is better in autumn and winter because there are no leaves on the trees. Therefore, the season can be used to improve the forecast in autumn and winter, to worsen in spring and summer. If forest areas are considered separately, the adjustment can be limited to these areas.

Reliability of individual measuring points: Depending on various factors, the reliability of a measuring point can be calculated. These

Reliability can be used in modeling to give individual measurements less weight than others or to ignore low-reliability measurements. The following influences can be used for the calculation:

o Measurement environment: building density, height of buildings, vegetation o Position accuracy

o Past measurements, with high variance the reliability is reduced

Additionally or alternatively, a self-learning model can be used, in which a multitude of additional parameters are collected, of which an influence on the achievable data rate is not yet known, but is possible. These measurements are correlated with measured data rates in order to find further influencing factors. Found influencing factors can be stored in the model in order to influence estimated data rates depending on these parameters.

In addition, the invention relates to a cover map, which was created according to the method with the features described above.

PREFERRED EMBODIMENTS OF THE INVENTION

The invention is explained in more detail below using an exemplary embodiment. They show in a purely schematic representation

Fig. 1 shows the sequence of the process for creating a

Cover map,

2 shows the principle of dividing passive measured values into groups,

3 shows an exemplary message format for a prediction

Request, and

Fig. 4 shows an exemplary message format for a provided

Prediction.

FIG. 1 shows the schematic sequence of the method 100 according to the invention. With the aid of a measuring fleet, which for example comprises at least one measuring vehicle 14 with a measuring device, for example a mobile communication means, active measurements are initially made to form the prediction model 11 17 performed. The measured active parameters are stored in a database 12. The measurement 17 is carried out to create the prediction model 11. The active measurements 17 can be repeated to update the measurement data. Then only measurements of passive parameters are necessary in order to then create the coverage map 10 based on the prediction model 11 and the passive parameters. The database 12, the prediction model 11 or the created coverage map 10 are located on a coverage map server 13.

If a coverage map 10 is required for a specific area,

For example, for points along a known route that is to be traveled by a vehicle 15 with a client device

Vehicle 15 made a prediction request 16, whereupon the

Prediction model 11 and measured passive parameters are used to create the coverage map 10 for the corresponding area along the travel route and to transmit the prediction 18 to the vehicle 15. The prediction 18 can be generated periodically or whenever new measurements have been carried out for a sub-area.

In FIG. 2, the principle of dividing passive measured values into groups is exemplified with two parameters and a division into two each

Value ranges for forming the prediction model 11 are shown. The range is given as an example, which would be used for the parameter combination SINR = -9 dB and RSRP = -121 dB. A first value range for the SINR parameter covers the values from -12 dB to 14 dB and a second value range covers the range greater than 14 dB to 40 dB. The first range of values for the RSRP parameter includes values from -140 dB to -92 dB and a second range includes values greater than -92 dB to -44 dB. Accordingly, the above parameter combination can be marked with the cross

Parameter group can be classified. For the model formation, data from different areas are preferably collected in order to achieve the best possible results

to get a general model.

FIG. 3 shows an exemplary message format for the prediction request 16 according to FIG. 2 by a vehicle 15 in JavaScript Object Notation (JSON). The prediction request is answered by searching for the point in the database 12 for each point of the known route in which this point lies. The prediction of the partial area stored in the database 12 is then added as a prediction for the position. The route is thus enriched with predicted data rates and then transmitted back to the requesting device as a prediction 18.

An exemplary message format in JSON for such a prediction 18 provided according to FIG. 2 is shown in FIG. In addition, the confidence value of the prediction 18 and further parameters can be included.

Claims

Expectations

1. A method (100) for creating a coverage map (10) for a wireless communication system, which comprises at least one network quality, among other things, wherein a prediction model (11) is created for predicting the network quality. wherein the network quality indicators are each assigned to a measuring point within the geographical area covered by the coverage map (10), at least one passive parameter being measured at a measuring point for calculating a network quality indicator, the measuring point using the prediction model (11) and the at least one passive parameter Parameters of the power quality indicator is calculated.

2. The method (100) according to claim 1, wherein for creating the

Prediction model (11) has at least one active one for each measuring point

Parameter, which is in particular a data rate, is measured.

3. The method (100) according to claim 1 or 2, wherein the prediction model (11) uses a machine learning method.

4. The method (100) according to any one of the preceding claims, wherein a passive parameter is a signal strength parameter.

5. The method (100) according to claim 1, wherein the geographical area covered by the cover map (10) is subdivided into sub-areas, each measuring point being assigned to a sub-area and being assigned from the measuring points to a sub-area

Power quality indicators, the median is calculated.

6. The method (100) according to any one of the preceding claims, wherein the

Measurement points are each grouped based on the passive parameter assigned to them.

7. The method (100) according to any one of the preceding claims, wherein the network quality indicator determined for each measuring point is stored in a memory (12).

8. The method (100) according to claim 7, wherein by means of the cover card (10)

the network quality indicators for waypoints in a known route are retrieved from the memory (12).

9. The method (100) according to any one of the preceding claims, wherein the measurement of the at least one passive parameter and the calculation of the

Power quality indicator is repeated cyclically at each measuring point or by an inquiry (16).

10. Cover map (10), created with a method according to one of the

Claims 1 to 9.