阅读说明：本技术 差分cqi值报告过程 (Differential CQI value reporting procedure ) 是由 斯特凡·托马辛 杨少石 马克·森特那罗 尼韦欧·本韦努托 于 2018-07-31 设计创作，主要内容包括：本公开涉及从终端设备到基站设备的CQI值反馈。本公开特别提出了差分CQI值报告过程,该过程在异常事件的情况下由CQI值预测支持。为此,本发明提出一种终端设备和一种基站设备。终端设备用于在当前时间间隔获得与基站设备和终端设备之间的DL信道相关的CQI值,基于获得的CQI值和先前时间间隔的CQI值计算差分CQI值,以及向基站设备发送差分CQI值。基站设备用于从终端设备接收差分CQI值,并且基于差分CQI值和先前时间间隔的CQI值获得当前时间间隔的CQI值。如果差分CQI值指示正常事件,则基站设备可以直接使用差分CQI值将先前时间间隔的CQI值更新为当前时间间隔的CQI值。如果差分CQI值指示异常事件,则基站设备可以例如通过机器学习算法从一个或多个先前时间间隔的CQI值预测当前时间间隔的CQI值。(The present disclosure relates to CQI value feedback from a terminal device to a base station device. The present disclosure proposes, among other things, a differential CQI value reporting procedure, which is supported by CQI value prediction in case of an abnormal event. To this end, the invention proposes a terminal device and a base station device. The terminal device is configured to obtain a CQI value related to a DL channel between the base station device and the terminal device at a current time interval, calculate a differential CQI value based on the obtained CQI value and a CQI value of a previous time interval, and transmit the differential CQI value to the base station device. The base station device is configured to receive the differential CQI value from the terminal device and obtain a CQI value for a current time interval based on the differential CQI value and a CQI value for a previous time interval. If the differential CQI value indicates a normal event, the base station apparatus may directly update the CQI value of the previous time interval to the CQI value of the current time interval using the differential CQI value. If the differential CQI value indicates an abnormal event, the base station apparatus may predict a CQI value for the current time interval from CQI values for one or more previous time intervals, e.g. by a machine learning algorithm.)

1. A terminal device (100) for

Obtaining a channel quality indicator, CQI, value (101) related to a downlink, DL, channel (102) between a network node (200) and the terminal device (100) at a current time interval,

calculating a differential CQI value (103) based on said obtained CQI value (101) and a CQI value (104) of a previous time interval, an

Sending the differential CQI value (103) to the network node (200).

2. Terminal device (100) according to claim 1, for

Obtaining the CQI value (101) within the determined subband and calculating the differential CQI value (103).

3. Terminal device (100) according to claim 1 or 2, wherein

The differential CQI value (103) is used to indicate:

-a state associated with a predefined change of the obtained CQI value (101) with respect to the CQI value (104) of the previous time interval, and

-indicating a state of an abnormal situation associated with an undefined change of the obtained CQI value (101) with respect to the CQI value (104) of the previous time interval.

4. Terminal device (100) according to any of claims 1 to 3, wherein

Each CQI value (101, 104) consists of four bits, and

the differential CQI value (103) consists of two bits.

5. A terminal device (100) according to claim 3 or 4, wherein the differential CQI value (103) is adapted to indicate an increase or decrease of the obtained CQI value compared to the CQI value of the previous time interval being larger than a predefined increase or decrease, the status indicating the abnormal situation.

6. Terminal device (100) according to claims 3 to 5, wherein

The differential CQI value (103) is used to indicate the following status:

-a predefined increase of the obtained CQI value (101) compared to the CQI value (104) of the previous time interval,

-a predefined reduction of the obtained CQI value (101) compared to the CQI value (104) of the previous time interval,

-the obtained CQI value (101) is unchanged compared to the CQI value (104) of the previous time interval,

-an increase or decrease of the obtained CQI value (101) compared to the CQI value (104) of the previous time interval being larger than the predefined increase or decrease, the increase or decrease being indicative of an abnormal situation.

7. Terminal device (100) according to any of claims 3 to 6, for

Storing the obtained CQI value (101) as the CQI value (201) of the current time interval if the differential CQI value (103) does not indicate an abnormal situation.

8. Terminal device (100) according to any of claims 3 to 6, for

Determining a predicted CQI value (300) from CQI values (104) of one or more previous time intervals if the differential CQI value (103) indicates an abnormal situation, an

Storing the predicted CQI value (300) as the CQI value (201) for the current time interval.

9. Terminal device (100) according to claim 8, configured to

The predicted CQI value is determined by executing a machine learning algorithm (300).

10. Terminal device (100) according to claim 9, for

Receiving one or more prediction coefficients from the network node (200), an

Executing the machine learning algorithm based on the received prediction coefficients.

11. A network node (200) for

Receiving a differential channel quality indicator, CQI, value (103) from a terminal device (100), wherein the differential CQI value (103) is related to a downlink, DL, channel (102) between the network node (200) and the terminal device (100), and

obtaining a CQI value (201) for a current time interval based on the differential CQI value (103) and a CQI value (104) for a previous time interval.

12. The network node (200) according to claim 11, wherein

The differential CQI value (103) is used to indicate:

-a status associated with a predefined variation of the CQI value (101) obtained by the terminal device (100) with respect to the CQI value (104) of the previous time interval, and

-a state indicating an abnormal situation associated with an undefined change of the obtained CQI value (101) of the terminal device (100) with respect to the CQI value (104) of the previous time interval.

13. The network node (200) of claim 12, configured to

Obtaining the CQI value (201) for the current time interval by updating the CQI value (104) for the previous time interval based on the change associated with the status indicated by the differential CQI value (103) if the differential CQI value (103) does not indicate an abnormal situation.

14. The network node (200) of claim 12, configured to

If the differential CQI value (103) indicates an abnormal situation, the CQI value (201) for the current time interval is obtained by determining a predicted CQI value (300) from CQI values (104) for one or more previous time intervals.

15. The network node (200) of claim 14, configured to

The predicted CQI value is determined by executing a machine learning algorithm (300).

16. The network node (200) of claim 15, further configured to

-sending one or more prediction coefficients obtained when executing the machine learning algorithm to the terminal device (100), in particular when the terminal device (100) enters a determined area (600) associated with the network node (200).

17. A method (400) for a terminal device (100), the method (400) comprising

Obtaining (401) a channel quality indicator, CQI, value (101) related to a downlink, DL, channel (102) between a network node (200) and the terminal device (100) at a current time interval,

calculating (402) a differential CQI value (103) based on said obtained CQI value (101) and a CQI value (104) of a previous time interval, an

Sending (403) the differential CQI value (103) to the network node (200).

18. A method (500) for a network node (200), the method (500) comprising

Receiving (501) a differential channel quality indicator, CQI, value (103) from a terminal device (100), wherein the differential CQI value (103) is related to a downlink, DL, channel (102) between the network node (200) and the terminal device (100), and

obtaining (502) a CQI value (201) for a current time interval based on the differential CQI value (103) and a CQI value (104) for a previous time interval.

19. A computer program product storing program code for performing the method according to claim 17 or 18 when the computer program is run by a processor.

Technical Field

The present disclosure generally relates to a process of feeding back Channel Quality Indicator (CQI) information from a terminal apparatus to a base station apparatus. The present disclosure proposes, in particular, a terminal device that reports to a base station device a differential CQI value related to a Downlink (DL) channel between the base station device and the terminal device. The differential CQI value reporting process of the present disclosure may be supported by CQI value prediction, which may be implemented at the base station device and the terminal device, respectively.

Background

In a conventional 3GPP frequency-division duplexing (FDD) cellular network, such as a network according to the Long Term Evolution (LTE) standard, CQI values are typically obtained by a User Equipment (UE) and reported to an evolved NodeB (eNB) associated with and communicating with it. The CQI value indicates the quality of the DL channel between the eNB and the UE. The eNB can select or adjust a Modulation and Coding Scheme (MCS) based on the received CQI values, e.g., the eNB may select an appropriate DL transmission rate for its next DL transmission to the UE on the DL channel.

The CQI value sent from the UE to the eNB is typically a 4-bit integer. The four "feedback bits" described above may cause transmission overhead, particularly if the UE obtains and reports individual CQI values for each of a plurality of subbands and/or if CQI values are reported from the UE to the eNB periodically and at short intervals. In wireless communication systems, transmission overhead is always a performance issue.

Disclosure of Invention

In view of this, the present disclosure is directed to improving the conventional CQI value reporting process. It is an object of the present disclosure to reduce transmission overhead, in particular by reducing the number of feedback bits reported from a terminal device (e.g. a UE) to a network node (e.g. to a base station device (e.g. an eNB)). The present disclosure is directed to achieving a fixed rate of CQI value reporting under stationary propagation conditions and to a procedure for coping with non-stationary conditions, e.g. in case of abnormal events. The present disclosure should operate without the network node locating the terminal device.

The object of the present disclosure is achieved by the solution provided in the appended independent claims. Advantageous embodiments of the present disclosure are further defined in the dependent claims.

In particular, the present disclosure proposes a process of reporting differential CQI values from a terminal device to a network node, wherein the reporting process can be supported by CQI value prediction at both devices. The CQI value prediction is implemented at the terminal device and the network node, respectively, such that the predicted CQI values are the same on both sides, i.e. such that the CQI value predictions are aligned.

A first aspect provides a terminal device for obtaining a CQI value relating to a DL channel between a network node and the terminal device at a current time interval, calculating a differential CQI value based on the obtained CQI value and a CQI value of a previous time interval, and transmitting the differential CQI value to the network node.

Since only the difference from the previous CQI value, i.e. the CQI value of the previous time interval, needs to be encoded, the differential CQI value is represented by fewer bits than the conventionally reported CQI value. Thus, the differential CQI value is represented by fewer bits than both the obtained CQI value and the CQI value of the previous time interval. Accordingly, transmission overhead caused by feedback of CQI information from the terminal device to the network node can be reduced, thereby improving network performance.

Herein, a time interval is a time period within which a specific CQI value is valid. The new time interval starts when a new CQI value is set, i.e. when the new CQI value takes effect. Thus, the CQI value for the current time interval is valid during the current time interval, while the CQI value for the previous time interval is valid in the previous time interval (i.e., before the CQI value for the current time interval is set). The lengths of the plurality of time intervals may be the same or different, i.e., new CQI values may be set periodically or aperiodically. Note that the obtained CQI value (obtained at the current time interval) refers to a measured or estimated value or a calculated CQI value, and is not necessarily the "CQI value for the current time interval". If the obtained CQI value is determined to be reliable, it may be set to the CQI value of the current time interval. Otherwise, for example, the obtained CQI value may not be considered, and the predicted CQI value may be set to the CQI value of the current time interval. This will be described in further detail below.

In one implementation of the first aspect, the terminal device is configured to obtain CQI values within the determined subbands and to calculate differential CQI values.

For example, the terminal device may obtain one differential CQI value for each of a plurality of subbands (particularly, for each subband included in the entire band). Thus, a higher feedback granularity may be achieved and at the same time feedback transmission overhead may be reduced. Optionally, the terminal device may also be configured to obtain CQI values and calculate differential CQI values within a wide band (i.e., within the entire band).

In another embodiment of the first aspect, the differential CQI value is used to indicate: a state associated with a predefined change of the obtained CQI value relative to the CQI value of the previous time interval, and a state indicating an abnormal situation associated with an undefined change of the obtained CQI value relative to the CQI value of the previous time interval.

The predefined change is an expected change of the obtained CQI value compared to the CQI value of the previous time interval, while the undefined change is a change that is not expected. The change may comprise an increase or decrease of the obtained CQI value relative to the CQI value of the previous time interval, or a constant CQI value.

The network node may deduce whether the obtained CQI value (e.g. the measured or estimated CQI value at the terminal device) is reliable or not by the status of the differential CQI value. That is, the network node knows whether the obtained CQI value represents a normal case or an abnormal case. The normal case is, for example, that the CQI value remains unchanged or that the CQI value increases/decreases slowly (i.e. predefined change), while the abnormal case is, for example, that the CQI value changes rapidly (i.e. undefined change).

Indicating an abnormal situation to a network node is an important part of the present invention. This is due to the fact that: the network node may thus decide how to obtain a CQI value for the current time interval based on the received differential CQI values, e.g. by updating the CQI value for the previous time interval according to the indicated change in the differential CQI value if no abnormal situation is indicated, or by other mechanisms yielding more reliable results if an abnormal situation is indicated. Such other mechanisms may be, for example, based on machine learning algorithms and/or CQI value prediction performed in a neural network (as will be described in detail below).

In another embodiment of the first aspect, each CQI value consists of four bits and the differential CQI value consists of two bits.

Thus, the feedback overhead can be reduced by as much as 50%. In particular, if such overhead reduction is obtained for each of a plurality of subbands, a substantial network performance gain will be achieved. Typically, each CQI value consists of more bits than the differential CQI value. In other words, each CQI value may be represented by M bits, a differential CQI value may be represented by N bits, and M > N. The above embodiments for the case where the CQI value is four bits and the differential CQI value is two bits are merely examples.

In a further embodiment of the first aspect, the differential CQI value is used to indicate an increase or decrease of the obtained CQI value compared to the CQI value of the previous time interval which is larger than a predefined increase or decrease, the above state indicating an abnormal situation.

In another embodiment of the first aspect, the differential CQI value is used to indicate the following status: a predefined increase of the obtained CQI value compared to the CQI value of the previous time interval, a predefined decrease of the obtained CQI value compared to the CQI value of the previous time interval, no change of the obtained CQI value compared to the CQI value of the previous time interval, an increase or decrease of the obtained CQI value compared to the CQI value of the previous time interval larger than the predefined increase or decrease (indicating an abnormal situation). For example, the predefined increase or decrease may be a smaller increase or decrease. An increase or decrease greater than a predefined increase or decrease may be indicated as a greater decrease.

In a smooth propagation condition (i.e. with the CQI value unchanged or only slowly varying) one of the normal states will be indicated. In non-stationary propagation conditions (e.g. in case the CQI value of the DL channel between the network node and the terminal device changes rapidly), an abnormal situation is indicated.

In another embodiment of the first aspect, the terminal device is configured to: if the differential CQI value does not indicate an abnormal situation, the obtained CQI value is stored as the CQI value of the current time interval.

That is, the CQI value measured or estimated by the terminal device may be set to the CQI value for the current time interval and remain valid until a new CQI value is set.

In another embodiment of the first aspect, the terminal device is configured to: if the differential CQI values indicate an abnormal situation, a predicted CQI value is determined from CQI values of one or more previous time intervals, and the predicted CQI value is stored as the CQI value of the current time interval.

In this case, after the differential CQI value is generated using the measured or estimated CQI value, the measured or estimated CQI value may be discarded. Instead, the predicted CQI value is set to the CQI value for the current time interval and remains valid until a new CQI value is set.

In particular, this prediction is used to support the proposed differential reporting procedure in case of non-stationary conditions (e.g. in case of an abnormal event) and is therefore an important part of the present invention. Since the CQI value obtained by the terminal device is not reliable in case of an abnormal event, CQI value prediction is beneficial for the proposed reporting procedure. This is notified to the network node by indicating an abnormal event in the differential CQI value and may trigger a CQI value prediction accordingly.

In another embodiment of the first aspect, the terminal device is configured to determine the predicted CQI value by executing a machine learning algorithm.

For example, the terminal device may execute a machine learning algorithm with a neural network. Particularly after providing the learning phase, the machine learning algorithm may ensure a high reliability of the CQI value prediction.

In another embodiment of the first aspect, the terminal device is configured to receive one or more prediction coefficients from the network node, and to execute the machine learning algorithm based on the received prediction coefficients.

Thereby, the prediction may be improved and the predicted CQI value may be further synchronized with the predicted CQI value obtained by the network node (separately). Thus, the CQI value prediction may always be aligned between the terminal device and the network node. This is an important part of the invention and enables the terminal device and the network node to continue the differential CQI value reporting process even in case of an abnormal event. Furthermore, the prediction process at the terminal device can be completed faster.

A second aspect provides a network node for receiving a differential CQI value from a terminal device, wherein the differential CQI value relates to a DL channel between the network node and the terminal device, and obtaining a CQI value for a current time interval based on the differential CQI value and a CQI value for a previous time interval. The network node may for example be a base station device.

Thereby reducing the transmission overhead resulting from the CQI information reporting procedure from the terminal device to the network node. Nevertheless, the network node may still obtain an accurate CQI value in order to e.g. adjust the MCS. Performance is not sacrificed while reducing transmission overhead. The network node may calculate a CQI value from the CQI value and the differential CQI value of the previous time interval and may set the calculated CQI value as the CQI value of the current time interval.

In one embodiment of the second aspect, the differential CQI value is used to indicate: a state associated with a predefined change of the CQI value obtained by the terminal device with respect to the CQI value of the previous time interval, and a state indicating an abnormal situation associated with an undefined change of the CQI value obtained by the terminal device with respect to the CQI value of the previous time interval.

The same advantages as described for the corresponding embodiments of the terminal device of the first aspect are achieved.

In another embodiment of the second aspect, the network node is configured to obtain the CQI value for the current time interval by updating the CQI value for the previous time interval based on a change associated with the state indicated by the differential CQI value if the differential CQI value does not indicate an abnormal situation.

In this case, since the CQI value obtained by the terminal device is reliable (because it reflects a normal situation), the network node directly uses the information in the differential CQI value to calculate and set the CQI value for the current time interval from the CQI value for the previous time interval.

In another embodiment of the second aspect, the network node is configured to obtain the CQI value for the current time interval by determining a predicted CQI value from CQI values for one or more previous time intervals if the differential CQI value indicates an abnormal situation.

In this case, since the CQI value obtained by the terminal device is not reliable (since it reflects an abnormal situation), the network node would rather predict the CQI value from the previously set (reliable) CQI value to compensate.

In another embodiment of the second aspect, the network node is configured to determine the predicted CQI value by executing a machine learning algorithm.

The same advantages as described for the corresponding embodiments of the terminal device of the first aspect are achieved.

In another embodiment of the second aspect, the network node is configured to send one or more prediction coefficients obtained when executing the machine learning algorithm to the terminal device, in particular when the terminal device enters a determined area associated with the network node.

Thus, the network node may synchronize its CQI value prediction with the CQI value prediction made by the terminal device. It can thus be ensured that the network node and the terminal device have the same CQI value at the current time interval. In fact, since the terminal device and the network node use the same prediction coefficients for the same machine learning algorithm, the terminal device and the network node remain synchronized at all times.

A third aspect provides a method for a terminal device, the method comprising: the method comprises obtaining a CQI value relating to a DL channel between the network node and the terminal device at a current time interval, calculating a differential CQI value based on the obtained CQI value and a CQI value of a previous time interval, and transmitting the differential CQI value to the network node.

In one embodiment of the third aspect, the method includes obtaining CQI values within the determined subbands and calculating differential CQI values.

In another embodiment of the third aspect, the differential CQI value is used to indicate: a state associated with a predetermined change in the obtained CQI value relative to the CQI value for the previous time interval, and a state indicating an abnormal situation associated with an undefined change in the obtained CQI value relative to the CQI value for the previous time interval.

In another embodiment of the third aspect, each CQI value consists of four bits and the differential CQI value consists of two bits.

In another embodiment of the third aspect, the differential CQI value is used to indicate the following status: a small increase of the obtained CQI value compared to the CQI value of the previous time interval, a small decrease of the obtained CQI value compared to the CQI value of the previous time interval, an unchanged obtained CQI value compared to the CQI value of the previous time interval, a large increase or decrease of the obtained CQI value compared to the CQI value of the previous time interval (indicating an abnormal situation).

In another embodiment of the third aspect, the method comprises: if the differential CQI value does not indicate an abnormal situation, the obtained CQI value is stored as the CQI value of the current time interval.

In another embodiment of the first aspect, the method comprises: if the differential CQI values indicate an abnormal situation, a predicted CQI value is determined from CQI values of one or more previous time intervals, and the predicted CQI value is stored as the CQI value of the current time interval.

In another embodiment of the first aspect, the method comprises: the predicted CQI value is determined by executing a machine learning algorithm.

In another embodiment of the first aspect, the method comprises: receiving one or more prediction coefficients from a network node; and executing a machine learning algorithm based on the received prediction coefficients.

The method of the third aspect and embodiments thereof achieve the same advantages and effects as described in relation to the terminal device of the first aspect and corresponding embodiments thereof.

A fourth aspect provides a method for a network node, the method comprising: receiving a differential CQI value from a terminal device, wherein the differential CQI value is related to a DL channel between a network node and the terminal device; and obtaining a CQI value for the current time interval based on the differential CQI value and the CQI value for the previous time interval.

In one embodiment of the fourth aspect, the differential CQI value is used to indicate: a state associated with a predefined change of the CQI value obtained by the terminal device with respect to the CQI value of the previous time interval, and a state indicating an abnormal situation associated with an undefined change of the CQI value obtained by the terminal device with respect to the CQI value of the previous time interval.

In another embodiment of the fourth aspect, the method comprises: if the differential CQI value does not indicate an abnormal situation, a CQI value for the current time interval is obtained by updating the CQI value for the previous time interval based on a change associated with the state indicated by the differential CQI value.

In another embodiment of the fourth aspect, the method comprises: if the differential CQI values indicate an abnormal situation, a CQI value for the current time interval is obtained by determining a predicted CQI value from CQI values for one or more previous time intervals.

In another embodiment of the fourth aspect, the method comprises: the predicted CQI value is determined by executing a machine learning algorithm.

In another embodiment of the second aspect, the method comprises: in particular, when the terminal device enters a certain area associated with the network node, one or more prediction coefficients obtained when executing the machine learning algorithm are sent to the terminal device.

The method of the fourth aspect and embodiments thereof achieve the same advantages and effects as described for the network node of the second aspect and corresponding embodiments thereof.

A fifth aspect provides a computer program product storing program code for performing the method according to the third and fourth aspects and embodiments thereof, when the computer program is run by a processor.

It should be noted that all devices, elements, units and means described in the present application may be implemented in software or hardware elements or any type of combination thereof. All steps performed by the various entities described in this application, as well as the functions described as being performed by the various entities, are intended to mean that the various entities are adapted or configured to perform the various steps and functions. Even if in the following description of specific embodiments the specific functions or steps performed by an external entity are not reflected in the description of specific detailed elements of the entity performing the specific steps or functions, it should be clear to a person skilled in the art that these methods and functions can be implemented in individual software or hardware elements or any type of combination thereof.

Drawings

The above aspects and embodiments of the invention will be explained in the following description of specific embodiments with reference to the drawings, in which

Fig. 1 shows a terminal device according to an embodiment.

Fig. 2 shows a network node (e.g., base station device) according to an embodiment.

Fig. 3 shows a reporting procedure performed by a terminal device according to an embodiment and a network node (e.g. a base station device) according to an embodiment. The terminal device and the network node form a system for performing a differential CQI value reporting procedure.

Fig. 4 shows a method for a terminal device according to an embodiment.

Fig. 5 illustrates a method for a network node (e.g., base station device) according to an embodiment.

Fig. 6 shows a terminal device according to an embodiment moving in a determined area associated with a network node (e.g. base station device) according to an embodiment.

Fig. 7 shows an example of the result of the proposed differential CQI value reporting procedure.

Fig. 8 shows the Cumulative Distribution Function (CDF) of the CQI value prediction error probability for three neural network configurations.

Detailed Description

Fig. 1 shows a terminal device 100 according to an embodiment. The terminal device 100 is specifically adapted to perform the differential CQI value reporting procedure as proposed herein. The terminal device 100 may be, for example, a UE in an LTE network. The terminal device 100 may be associated to a network node, such as a base station device 200 (see also fig. 2), for example to an eNB in an LTE network. The terminal apparatus 100 may receive data from the base station apparatus 200 via the DL channel 102, and may transmit data to the base station apparatus 200 via an Uplink (UL) channel (not shown). Hereinafter, for the sake of brevity, the description of the embodiments will refer to a base station apparatus.

The terminal device 100 is specifically configured to obtain a CQI value 101 related to a DL channel 102 between the base station device 200 and the terminal device 100 at a current time interval. For example, the terminal device 100 may estimate the DL channel 102 for this purpose, and may calculate the above-described CQI value 101 from a signal-to-noise ratio (SNR). In a normal case, the terminal device 100 may store the obtained CQI value 101 as a CQI value for the current time interval. In an abnormal case (described later), the terminal device 100 may ignore the obtained CQI value 101, may predict a CQI value, and may store the predicted CQI value as a CQI value of the current time interval.

Furthermore, the terminal device 100 is configured to calculate a differential CQI value 103 based on the obtained CQI value 101 and the CQI value 104 of the previous time interval. To this end, the terminal device 100 may retrieve the CQI value 104 of the previous time interval from a memory or the like. Thus, the terminal device 100 may be configured to store one or more CQI values 104 for one or more previous time intervals and may store the CQI value for the current time interval immediately upon setting the CQI value (i.e. when a final decision is made on the CQI value for the current time interval). It should be noted that the terminal device 100 generates the differential CQI value 103 using the obtained CQI value 101 regardless of whether the obtained CQI value 101 is set to the CQI value of the current time interval.

Further, the terminal apparatus 100 is configured to transmit the differential CQI value 103 to the base station apparatus 200. The terminal device 100 may specifically report the differential CQI value 103 by the same mechanism as a legacy terminal device (e.g., UE) reports a legacy full CQI value.

Fig. 2 shows a base station apparatus 200 according to an embodiment of the present invention. The base station apparatus 200 is configured to perform the differential CQI value reporting procedure proposed herein. The base station apparatus 200 may be, for example, an eNB in an LTE network. The base station device 200 may be associated to at least one terminal device 100 (see also fig. 1) (e.g. a UE in an LTE network). The base station apparatus 200 may receive data from the terminal apparatus 100 via an UL channel (not shown), and may transmit data to the terminal apparatus 100 via the DL channel 102.

The base station device 200 is specifically configured to receive a differential CQI value 103 from the terminal device 100, wherein the differential CQI value 103 is related to the DL channel 102 between the base station device 200 and the terminal device 100. The base station device 200 may specifically receive the differential CQI value 103 by the same mechanism as a legacy base station device (e.g. eNB) receiving a legacy full CQI value from a legacy terminal device.

Further, the base station apparatus 200 is configured to obtain a CQI value 201 for a current time interval based on the differential CQI value 103 and the CQI value 104 for a previous time interval. Specifically, the base station apparatus 200 can calculate the CQI value 201 of the current slot directly from the CQI value 104 of the previous time interval using the differential CQI value 103. This is especially true under normal circumstances. The base station apparatus 200 may also decide to prefer a predicted CQI value based on the differential CQI value 103 and set the predicted CQI value as the CQI value of the current time interval.

Fig. 3 shows a proposed differential CQI value reporting procedure performed by a terminal device 100 according to an embodiment of the present invention (based on the terminal device 100 of fig. 1) and a base station device 200 according to an embodiment of the present invention (based on the base station device 200 of fig. 2). Like elements in fig. 1, 2, and 3 are labeled with like reference numerals and function identically.

For the differential CQI value reporting process, the full CQI value is assumed to have four bits, i.e. CQI [ n ] ∈ {0, 1.. 15} is assumed to be the value of the 4-bit quantized CQI calculated by the terminal device 100 at time interval n, and can take 16 different values (or units). It is proposed that the differential CQI value 103 has only two bits, i.e. only half the bits of the full CQI value. In time interval n, the terminal device 100 reports to the base station device 200 two bits, wherein the two bits describe the difference between the CQI value 101 obtained by the terminal device 100 for the current interval and the CQI value 104 valid for the previous time interval n-1. Four cases can be indicated with such a two-bit differential CQI value 103:

CQI [ n ] ═ CQI [ n-1] + x (e.g., x ═ 1), i.e., the obtained CQI value 101 is increased by a first predefined amount, which is a smaller amount x than the previous CQI value 104. For example, one of the above 16 possible units is increased (this event may be associated with bit pair "10" of differential CQI value 103).

CQI [ n ] ═ CQI [ n-1] -x (e.g., x ═ 1), i.e., the obtained CQI value 101 is reduced by a second predefined amount, which is a smaller amount x than the previous CQI value 104. For example, by one unit (the event may be associated with bit pair "01").

CQI [ n ] ═ CQI [ n-1], i.e. the obtained CQI value 101 remains the same as the previous CQI value 104 (this event may be associated with bit pair "00").

| CQI [ n ] -CQI [ n-1] | > x (e.g., x ═ 1), i.e., the obtained CQI value 101 changes by a larger amount (greater than x) than the previous CQI value 104. The larger amount is larger than the first predetermined amount and/or the second predetermined amount as defined above. For example, more than one unit is changed (the exception event may be associated with bit pair "11").

It should be noted that the initialization phase is preferably applied to start the reporting process proposed by the present invention. The base station apparatus 200 and the terminal apparatus 100 may start from the normalized CQI value and then may use the differential method until the correct value is reached.

Limited variability of CQI values over a short period has been demonstrated in the literature and is also required by 3GPP to maximize the continuity of MCS. This means that in the above exemplary procedure, most of the differential CQI values 103 will indicate one of the first three states (i.e., -1, 0, or +1) of the obtained CQI value 101 compared to the previous CQI value 104, compared to the previous CQI value 104. However, the attenuation seen by the terminal device 100 may vary depending on its position in a determined area, also referred to as region of interest (ROI), associated with the base station device 200. For example, the distance between the terminal apparatus 100 and the base station apparatus 200 may change and/or the propagation condition may change. Specifically, an abnormal event ("11" in the above-described exemplary example of the differential CQI value 103) may occur, for example, at street corner or at a transition between a line-of-sight (LOS) propagation condition and a non-LOS (NLOS) propagation condition (or vice versa).

In the case where an abnormal event indicated by the differential CQI value 103 occurs, the base station apparatus 200 is not suggested to calculate the CQI value 201 of the current time interval directly from the previous CQI value 104 and the differential CQI value 103. But rather suggests that the base station apparatus 200 predicts a CQI value 201 for the current time interval from one or more CQI values 104 for the past time interval. In particular, a machine learning method may be used for the prediction as follows.

During the training phase, the machine learning algorithm (or the entity performing the machine learning algorithm) may learn the correspondence between the CQI value before the abnormal event and the missing CQI value, in particular using the legacy 4-bit CQI feedback of LTE. During the development phase, upon receiving an indication of an abnormal event from the terminal device 100, the machine learning algorithm may provide a predicted CQI value 300 to the base station device 200 based on observations of past CQI values (see "CQI predictor" in fig. 3).

The proposed protocol may lead to a degradation of system performance if the predictions made by the machine learning algorithm are erroneous. In this case, when the differential process is restarted after the abnormal event, the base station apparatus 200 will be misaligned with the terminal apparatus 100. To prevent this, the base station apparatus 200 may transmit the prediction coefficients of the machine learning algorithm to the terminal apparatus 100 (particularly when the terminal apparatus 100 enters the ROI). The terminal device 100 may use the same machine learning algorithm as the base station device 200 to predict the CQI value 201 for the current time interval from the CQI values 104 for one or more previous time intervals, and the two devices 100 and 200 may remain fully synchronized due to the prediction coefficient 304.

Therefore, at the time of an abnormal event, the base station apparatus 200 and the terminal apparatus 100 can obtain the predicted CQI value 300 (represented by CQIpred [ n ]) using, for example, a neural network. Thus, the differential CQI value 103 used for reporting at the next time interval n +1 is CQI [ n +1] -CQIPred [ n ], and the differential scheme may be applied as described above.

The operational flow of fig. 1 can be summarized as follows. On the terminal device 100 side, the terminal device 100 obtains a CQI value 101 for the current time interval, for example, by measuring or estimating a DL channel 102 and calculating the CQI value 101 from a signal-to-noise ratio (SNR), for example. Then, the terminal apparatus 100 feeds back the difference between the obtained CQI value 101 and the previous CQI value 104 (stored in a local memory, for example) to the base station apparatus 200.

For the differential CQI value 103, 2 bits are used. If the differential CQI value 103 is, for example, "11" (indicating an abnormal event) as determined at block 303, the terminal device 100 determines a predicted CQI value 300 from the CQI values 104 of one or more previous time intervals and updates the memory with the predicted CQI value 300. Alternatively, the CQI value 104 of the previous time interval in memory is updated to the CQI value 201 of the current time interval, taking into account the differential CQI value 103. Or simply stores the obtained CQI value 101 as the CQI value 201 for the current time interval.

On the base station apparatus 200 side, the base station apparatus receives a feedback message containing CQI correction (i.e., differential CQI value 103) from the terminal apparatus 100. If the differential CQI value 103 is "11" (indicating an abnormal event) as determined at block 302, the base station apparatus 200 activates a machine learning entity ("CQI predictor") to predict the CQI value 300 and provides it to, for example, an MCS selection block 301. That is, the base station apparatus 200 obtains a CQI value 201 for a current time interval by determining a predicted CQI value 300 from CQI values 104 for one or more previous time intervals. If the differential CQI value 103 is not "11", for example, the base station apparatus 200 supplies the updated CQI value 201 to the MCS selection block 301. That is, the base station apparatus 200 obtains the CQI value 201 of the current time interval by updating the CQI value 104 of the previous time interval based on the change associated with the state indicated by the differential CQI value 103.

Fig. 4 shows a general method 400 for the terminal device 100 according to an embodiment of the invention according to the above. The method 400 includes step 401: a CQI value 101 associated with a DL channel 102 between the base station apparatus 200 and the terminal apparatus 100 is obtained at the current time interval. Further, the method 400 comprises step 402: calculating a differential CQI value 103 based on the obtained CQI value 101 and the CQI value 104 of the previous time interval, and step 403: the differential CQI value 103 is transmitted to the base station apparatus 200.

Fig. 5 shows a general method 500 for a base station apparatus 200 according to an embodiment of the invention according to the above. The method 500 comprises the steps 501: a differential CQI value 103 is received 501 from the terminal device 100, wherein the differential CQI value 103 is related to the DL channel 102 between the base station device 200 and the terminal device 100. Further, the method 500 comprises step 502: a CQI value 201 for the current time interval is obtained based on the differential CQI value 103 and the CQI value 104 for the previous time interval.

Exemplary embodiments are set forth below. In this embodiment, referring to fig. 6, the determined region 600 is considered to be an ROI, in particular a square region with sides S having manhattan topology with roads 602 and building blocks 601. The base station apparatus 200 is located at the center of the area 600 (e.g. at the top of the mast at the center of a roundabout with a radius of 5 m). Assume that the terminal device 100 makes multiple passes through a randomly generated path 603 along a street 602 in the area 600, each time changing the selected street lane and speed. To sample such a shifting pattern, the time axis is divided into intervals of length τ (which may be equal to the CQI reporting period), i.e., τ -N_pdT_TTIWherein, T_TTIIs the duration of the transmission time interval.

In summary, embodiments based on determining the region 600 illustratively assume that:

manhattan topology

Side length of 525m

Road width of 15m

Block size 75m

The base station apparatus 200 is located at the center of the area 600

Terminal device 100 repeats the same path multiple times (changing street lanes and speed)

·N_pd＝160

The striped area 604 is in LOS, the others are in NLOS conditions.

The wireless channel model is assumed as follows. As in most literature, the DL radio channel model is viewed as having a global gain caused by:

"propagation loss": differentiating LOS propagation from NLOS propagation

"correlated shading": lognormal distribution of random variables due to the presence of buildings

The "fast-fading" component, due to multipath propagation and following the rayleigh distribution, can be ignored (since it is under consideration (N)_pd160) averaged on a time scale)

The radio channel parameters are summarized in the following table:

fig. 7 depicts example results of the reporting scheme presented above. Specifically, the method comprises the following steps:

the top graph shows the SNR time series

The diagram in the middle shows the CQI time series

The bottom graph shows the proposed differential CQI feedback, where the feedback value "2" usually only represents an abnormal situation.

It can be seen that most traces can be described efficiently by the envisaged differential CQI value reporting scheme, except for a few samples that are in an abnormal situation due to LOS-NLOS conversion (and vice versa). In this case, the proposed machine learning method would provide missing CQI values. It is observed that even if the predictions made by the machine learning entity are not optimal, the base station apparatus 200 and the terminal apparatus 100 remain synchronized, since the base station apparatus 200 (particularly when the terminal apparatus 100 enters the area 600) transmits prediction coefficients of the machine learning algorithm to the terminal apparatus 100, and it is possible to periodically update these prediction coefficients.

The performance of the differential feedback process is evaluated. In particular, to obtain performance statistics for the process, the operation of the process was simulated on Q-50 randomly generated paths, each path having Q-250 copies (with different travel speeds in the range [2.5, …, 20] m/s). Note that only paths without loops are considered.

The machine learning algorithm is implemented by a neural network with 1 hidden layer of size γ. To implement a neural network, a pattern recognition network is employed that is trained to classify inputs according to target classes. In the present case, the input to the neural network is the CQI value of P non-consecutive samples before the abnormal event. There are two reasons for using CQI values spaced apart by μ intervals: on one hand, the complexity of the neural network is reduced; on the other hand, neural networks are more robust to speed up changes between different implementations. The goal of the neural network is a CQI value corresponding to an abnormal event. In the development phase, CQI tracking is provided, which is different from the tracking provided for the training phase. P_errDefined as the probability of a CQI value prediction error in the event of an observed anomalous event.

FIG. 8 shows P for various neural network configurations_errA Cumulative Distribution Function (CDF). For the training phase and the development phase, a sequence of CQIs down sampled by a factor μ ═ 2 is used. It can be seen that the neural network is always correct in about 20% of the randomly generated paths, therefore, the proposed difference method is error free and effectively reduces the feedback to 1/2. On the other hand, considering all the simulated paths, in the worst case, at most about 25% of the neural network prediction errors. It can also be seen that the error rate is lower than 5% (15%) in 50% (85%) of these paths. It should also be noted that the performance gap between neural networks having different configurations is negligible. It can also be observed from simulations that neural network errors are generally due to similarity of CQI sequences before (i.e. with different CQI targets) different exceptional events. In these cases, the neural network may mistake one event for another, resulting in a wrong prediction of the CQI value.

The disclosure has been described in connection with various embodiments as examples and implementations. However, other variations will be understood and effected by those skilled in the art in practicing the claimed subject matter, from a study of the drawings, the disclosure, and the independent claims. In the claims and specification, the word "comprising" does not exclude other elements or steps, and the indefinite article "a" or "an" does not exclude a plurality. A single element or other unit may fulfill the functions of several entities or items recited in the claims. The mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be used to advantage.