Apparatus and method for determining transmission policies for a plurality of applications of different types

Technical field of equipment and method for determining transmission strategy for multiple different types of applications

[0001] The present invention relates to the field of optimizing wireless network structure and resource allocation.

Background technique

[0002] The optimization of the network structure is key to achieving the maximum network capacity and providing high-quality services for the maximum possible number of users. In a typical scenario, multiple users share wireless media and perform very different applications, such as video, voice, and FTP transmission. Optimizing resource allocation across all users and applications will maximize user satisfaction.

[0003] So far, cross-layer optimization has only been applied to a single application system. However, in practice, multiple users sharing a wireless medium (such as in a cell) often run different applications at the same time. User satisfaction is transformed into a collection of different needs for each type. In addition, the impact of loss on the user's perceived quality is highly dependent on the application.

[0004] The problem of optimizing across multiple applications has been mainly dealt with in the form of throughput maximization, such as V. Tsibonis, L. Georgiadis, L. Tassiulas in Exploiting wireless channel state information for throughput maximization IEEE INFOCOM 2003 (hereinafter referred to as [TsiOl]), and by Xin Liu, E. Chong, N. Shroff, in "Transmission scheduling for efficient wireless ut subscription IEEE INFOCOM 2001 (hereinafter referred to as [LiuOl]).

[0005] Throughput maximization makes it possible to provide optimal performance only for applications that are not sensitive to delay and packet loss. Multimedia applications, such as video and voice, are highly sensitive to changes in data rate, delay, and packet loss. Even the importance of grouping changes dynamically based on the history of previous groupings. For these reasons, maximizing throughput results in performance that is generally not optimal relative to the quality of multimedia applications perceived by users.

[0006] WO 00/33511A discloses a system for improving the quality of service for end users in a packet-switched network. Reports are sent from various nodes in the network to inform the network administrator of the end user's service quality at that node, which represents a quality estimate based on the end user's perception. The quality manager analyzes the report and sends a command to the node, which sends the report and/or to other nodes to improve the end user service quality at the node and in the packet-switched network as a whole. Nodes include sending and receiving terminals, routers and gateways. The report includes measurements of link parameters, device parameters, and end-user service quality.

Summary of the invention

[0007] The purpose of the present invention is to provide a device and method for determining the transmission strategy of a plurality of different types of applications in consideration of the user perception quality of the applications.

[0008] The object is to adopt a device for determining transmission strategies for a plurality of different types of applications according to claim 1, a method for determining transmission strategies for a plurality of different types of applications according to claim 15, and a computer according to claim 16. Program implementation.

[0009] The present invention provides a device for determining a transmission strategy for a plurality of different types of applications, the determination is based on

[0010] the first transmission data associated with the first transmission of the first transmission type,

[0011] The second transmission data associated with the second transmission of the second transmission type,

CN 101026552 Β

[0012] The device includes:

[0013] means for obtaining a first score in a shared range, the first score being based on an evaluation of the first transmission data in a specific manner of a first transmission type;

[0014] a device for obtaining a second score in a shared range, the second score being based on an evaluation of the second transmission data in a second transmission type specific manner;

[0015] The means for determining, based on the obtained score, determines respective first and second transmission strategies for the first and second transmissions, and each transmission strategy defines one or more transmission parameters so that all The sum of the first expected score and the second expected score is the largest.

[0016] In addition, the present invention provides a method for determining a transmission strategy for a plurality of different types of transmissions, the determination is based on

[0017] the first transmission data associated with the first transmission of the first transmission type,

[0018] the second transmission data associated with the second transmission of the second transmission type,

[0019] The method includes the following steps:

[0020] obtaining a first score in the shared range, the first score being based on an evaluation of the first transmission data in a specific manner of the first transmission type;

[0021] obtaining a second score within the shared range, the second score being based on an evaluation of the second transmission data in a second transmission type specific manner;

[0022] Based on the allocated first and second scores, the respective first and second transmission strategies are determined for the first and second transmissions, and each transmission strategy defines one or more transmission parameters such that for subsequent transmission intervals The sum of the first expected score and the second expected score of is the largest.

[0023] In addition, the present invention provides a computer program with program code, and when the program runs on a computer, the program code executes the method according to claim 15.

[0024] The present invention is based on the following decision, which collectively optimizes the system for different user and application needs:

[0025] First, a common metric is defined, which quantifies user satisfaction with service delivery, and second, network and/or application parameters are mapped to the metric.

[0026] In this context, the shared metric is also referred to as a score, where the score is defined for a shared range, and the shared range has a shared minimum score and a shared maximum score, where the sharing is used for all transmissions and transmissions. Defined in the sense of type sharing.

[0027] The present invention provides a cross-layer optimization framework, the purpose of which is to maximize user satisfaction. The difficulty of the method of the present invention lies in the problem of quantifying user satisfaction with respect to system parameters, such as throughput, delay, and packet error rate.

[0028] In a preferred embodiment, a mean opinion score (MOS) is used as the score and shared performance metric for optimization. Although the following discussion will be based on the average opinion score (MOS), it should be noted that the present invention includes the use of other scores; its scores with a common range different from the average opinion score (MOS) are also possible, and other scores take into account the transmission of user perception quality.

[0029] The Mean Opinion Score (MOS) was originally proposed for speech quality evaluation, and provides a digital measurement of human voice quality at the destination end of the circuit. This scheme uses multiple subjective tests (arbitrary scores), and mathematically averages these subjective tests to obtain a quantitative indicator of system performance. To determine the average opinion score (MOS), multiple listeners rate the quality of test sentences read aloud by the speaker on the communication circuit. The listener gives each sentence the following grades: (1) bad; (2) poor; ; average (4); good; excellent. Mean Opinion Score (M0S) is all individual scores

CN 101026552 Β

Arithmetic average, and can range from 1 (worst) to 5 (best).

[0030] For other applications, such as video, web browsing, and file download, the average opinion score (MOS) scale that reflects the application quality perceived by the user is used. In this way, this enables the use of a common optimization metric to optimize all applications. The objective function can be selected, for example, as the average opinion score (M0S) of all users or all transmissions:

[0031] = ⑴

K k=\

[0032] Where F@). is the objective function with a cross-layer parameter element MxeX. Product is a collection of all possible parameter tuples extracted from the protocol layer.

[0033] Wk is the relative importance of the user or transmission as determined by the service agreement between the user and the service provider.

[0034] The decision of the device for determining (hereinafter also referred to as the optimizer) can be expressed as follows:

[0035] x_opt = arg max F(x)

[0036] χ _e J

[0037] where is the optimal parameter tuple, which maximizes the objective function. Once the optimizer has selected the optimal values of the parameters, it publishes them to all the individual layers, which are responsible for converting them back to the actual operating mode.

[0038] Using scores with a common scale or a common range-with the same minimum score and the same maximum score-as optimization parameters provides various advantages. First, since the user-perceived service or application quality is provided on the same scale or range, the scale or range is common to all transmission or application types. Therefore, in addition to using different channel codecs and different modulation schemes, for example, In addition to taking advantage of the differences at the physical layer, it is also possible to take advantage of the differences at the application layer by, for example, using different source codecs (codec=encoding/decoding). Therefore, it is possible to calculate all possible transmission scenarios including all applications and their possible transmission parameters, and compare them with each other based on the specific scores of the transmission scenarios. The scene-specific score may be the sum of all the "single transmission" scores or the arithmetic average of all the "single transmission" scores. Thus, the task of the device or optimizer used for determination is to treat all applications or transmissions "equally" and maximize the sum or arithmetic average of all individuals. For example, the operation of extracting application and physical layer parameters for the user's perceived quality score provides an effective way to optimize network or wireless resource allocation, and at the same time takes into account the real-time and latency requirements of applications such as voice and video streaming. Second, for example, using scores in a common range facilitates prioritizing specific users or applications and/or providing a fair distribution of network and wireless resources based on the score history of each application or user.

[0039] The basic cross-layer optimization method, the principle of parameter extraction for multi-user cross-layer optimization, and the formulation of the objective function are explained in more detail by Y. Peng, S. Khan, E. Steinbach, M. Sgroi, W. Ke Iler er in Adaptive resource allocation and frame scheduling for wireless multi-uservideo streaming^, IEEE International Conference on Image Processing, ICIP, 05, Genova, Italy, September 2005 (hereinafter referred to as [PenOl]) and by S. Khan, M. Sgroi, E. Steinbach and W. Kellerer is provided in Cross-layer optimization for wireless videostreaming-performance and cost" IEEE International Conference on Multimedia & Expo, ICME 2005 Amsterdam, July 2005 (hereinafter referred to as [KhaOl]).

Description of the drawings

[0040] The preferred embodiments of the present invention will be described in detail with reference to the following drawings:

CN 101026552 Β

[0041] FIG. 1A shows an embodiment of the inventive device;

[0042] FIG. 1B shows an exemplary network scenario with a base station including the inventive device;

[0043] FIG. 1C shows a diagram for multi-application, cross-layer optimization;

[0044] FIG. 1D shows a diagram explaining the relationship between mean opinion score (MOS) and user satisfaction;

[0045] FIG. 2 shows a diagram describing the perceptual speech quality evaluation (PESQ) based on the mean opinion score (MOS) relative to packet loss of different speech codecs;

[0046] FIG. 3 shows a graph of File Transfer Protocol (FTP) User Average Opinion Score (MOS) estimation surface relative to packet loss and data range;

[0047] FIG. 4A shows an exemplary H.264-based encoding of a video sequence for a conversational video application;

[0048] FIG. 4B shows a graph describing the average opinion score (MOS) of a video user relative to the peak signal-to-noise ratio (PSNR); [0049] FIG. 4C shows a graph describing the average opinion score (MOS) of a video user relative to the foreman video A picture with missing slices of the sequence;

[0050] FIG. 5A shows a diagram of the simulation architecture of the present invention;

[0051] FIG. 5B shows an embodiment of the inventive method for the simulation architecture shown in FIG. 5A;

[0052] FIG. 6 shows a graph of the average opinion score (MOS) of a voice user based on the simulation architecture of FIG. 5A; [0053] FIG. 7 shows a file transfer protocol (FTP) user based on the simulation architecture of FIG. 5A A graph of the average opinion score (MOS);

[0054] FIG. 8 shows a graph of the average opinion score (MOS) for a video conference according to the simulation architecture of FIG. 5A; [0055] FIG. 9 shows the use of a system at a system symbol rate of 500k symbols/sec. The average opinion score (MOS) gain per user based on the simulation architecture of FIG. 5A;

[0056] FIG. 10 shows a graph of the average opinion score (MOS) gain per user according to the simulation architecture of FIG. 5A at a system symbol rate of 700k symbols/sec; and

[0057] FIG. 11 shows a graph of the average opinion score (MOS) gain per user according to the simulation architecture of FIG. 5A at a system symbol rate of 900k symbols/sec.

Detailed ways

[0058] FIG. 1A shows an embodiment of the inventive device 100, which includes: a device 110 for obtaining a first score in a common range, a device 120 for obtaining a second score in a common range, and An apparatus 130 for determining a transmission strategy. The device 110 is configured to obtain a first score 112 within a common range, the first score 112 being based on an evaluation of the first transmission data associated with the first transmission of the first transmission type, where the first transmission type is performed in a specific manner The evaluation. The device 110 for obtaining the first score 112 outputs the first score 112 to the device 130 for determining. The device 120 for obtaining a second score within the common range is operable to obtain the second score 122 and output the second score 122 to the device for determination 130, wherein the second score 122 is based on the An evaluation of the second transmission data associated with the second transmission of the second transmission type, wherein the evaluation is performed in a manner specific to the second transmission type.

[0059] The device 130 for determining is operable to receive the first score 112 and the second score 122, and determine the respective first and second transmissions based on the scores 112 and 122. A and a second transmission strategy 132, where each transmission strategy defines one or more transmission parameters to maximize the sum of expected scores.

[0060] In an embodiment of the inventive device 100, the means 110, 120 for obtaining are operable to use the previous most

CN 101026552 Β

One or more of the expected scores in the step of increasing, obtain the first or second score. In this case, the device is suitable for operation without feedback from other entities (such as receivers).

[0061] In another embodiment of the inventive device 100, the means 110 and 120 for obtaining respectively include means 114 and 124 for receiving the first score 112 from the device to which the first or second transmission data is sent. Or the second score 122, where the receiving means 114, 124 are optional, and are shown in dotted lines in FIG. 1A.

[0062] In yet another embodiment of the inventive device 100, the means for obtaining 110.120 includes respective means 116 and 126 for obtaining the obtained results from the first transmission characteristic 118 or the second measured transmission characteristic 128. The first score 112 or the second score 122, the first transmission characteristic 118 or the second measured transmission characteristic 128 may be received, for example, from the device to which the first or second transmission data is sent, wherein the means for receiving 116.126 is optional and is shown in dashed lines in Figure 1A. For example, the measured transmission characteristics 118, 128 may be a transmission rate, a signal-to-noise ratio (SNR), or a packet error rate (PEP), where, for example, the packet error rate may be evaluated based on the signal-to-noise ratio.

[0063] In an alternative embodiment of the inventive device 100, the devices 110, 120, and 130 may be combined into one device.

[0064] An alternative embodiment of the inventive device 100 may include more than two means 110.120 for obtaining scores, wherein the means for obtaining is operatively based on transmitting data in a third, fourth, etc. application type specific manner. In the evaluation, get third and fourth scores. Transmission data can be associated with third, fourth, etc. transmission types.

[0065] Another embodiment of the inventive device 100 may not include two independent devices 110, 120, but a shared device for obtaining, wherein the sharing device for obtaining is operatively based on the first or second The transmission type specific method evaluates the transmission data and obtains the first or second score. The first or second transmission type specific method depends on whether the transmission data is associated with the first transmission type or the second transmission type.

[0066] FIG. 1B shows an exemplary network scenario with a base station 160, a first terminal 170, and a second terminal 180. The base station 160 includes the invention device 100 and an antenna 162 connected to the invention device 100.

[0067] FIG. 1B shows a scenario in which the first transmission data associated with the first transmission of the first transmission type is between the first terminal 170 and the base station 160 in the downlink 170D or the uplink 170U or Transmission on both. Therefore, the second transmission data associated with the second transmission of the second transmission type is transmitted between the second terminal 180 and the base station 160 on the downlink 180D or the uplink 180U or both.

[0068] The following will discuss different scenarios for obtaining the score based on the transmission between the first terminal 170 and the base station 160, and the following explanation can also be applied to the second terminal 180 and the base station 160 or any The second transmission between other terminals.

[0069] In a downlink scenario, the base station 160 sends the first transmission data to the first terminal 170 in the downlink 170D±. The first terminal 170 receives the first transmission data, and can measure, for example, the actual signal-to-noise ratio (SNR) of the received transmission data.

[0070] For example, the first terminal 170 is operable to transmit a transmission of the received first transmission data to the base station, or more precisely, to the device 110 for obtaining the first score, on the uplink 170U± Characteristic, that is, the actual signal-to-noise ratio of the received first transmission data, or the first score itself is operatively obtained from the first transmission characteristic, and sent to the base station 160 on the uplink 170U or used to obtain The first scoring device 110 sends the first score.

[0071] In the uplink scenario, the first transmitter 170 sends the first transmission data to the base station 160 on the uplink 170U. The base station 160, or more precisely, the obtaining device 110 may determine the transmission characteristic of the first transmission data by itself, and directly obtain the first score 112 therefrom.

[0072] Based on the obtained scores, for example, the first and second scores 112, 122, the device 130 determines the respective first and second scores.

CN 101026552 Β

The second transmission strategy (132.1; 132.2) defines the one or more transmission parameters for the first and second transmissions, so that for transmissions in subsequent transmission intervals, the first expected score and the The sum of the second expected scores is the largest.

[0073] For the aforementioned downlink transmission scenario 170D, the base station 160 will send the first transmission data in the subsequent transmission interval based on the determined transmission parameters.

[0074] For the latter scenario, that is, the uplink scenario 170U, the base station 160 sends the determined first transmission strategy to the first terminal 170 on the downlink, and then the terminal will be based on the reception from the base station 160 In the first transmission strategy, the first transmission data is sent in consecutive transmission intervals.

[0075] For example, a dedicated signaling channel may be used to send the transmission characteristic from the first terminal 170, or, for example, the transmission characteristic may be piggybacked into the confirmation message.

[0076] The term transmission data associated with the transmission of a certain transmission type includes the transmission of application data, such as voice, hypertext transfer protocol (HTTP), file transfer protocol (FTP), video and music streaming, and other applications, But it also includes the transmission of signaling data or any other data (for example, data used to control the network).

[0077] In a typical wireless or mobile network scenario, base stations control the wireless resources in their cells. Therefore, the inventive device used to determine transmission strategies for multiple different types of applications is typically in the base station. achieve. However, in alternative scenarios, like in an ad-hoc network, any other device, such as a communication device, can manage tasks and/or optimize wireless resources. For these situations, the inventive device can be implemented in other devices for optimizing or maximizing user-perceived service and application quality.

[0078] As shown in FIG. 1C, the following shows that the optimized architecture of the present invention achieves important improvements in terms of user perception quality for an example with three application types, the three application types being real-time voice, file download And video conferencing.

[0079] FIG. 1C shows a diagram for an exemplary multi-application, cross-layer optimization architecture, which includes an inventive device 100, referred to as a cross-layer optimizer in FIG. 1C, which receives as a transmission from the wireless link layer Characteristic transmission characteristics 118,128, such as transmission rate, packet error rate (PEP) and/or packet size.

[0080] The cross-layer optimizer, namely the inventive device 100, is operable to obtain the score from the transmission characteristics based on a predefined lookup table or a predefined algorithm, that is, the average opinion score (MOS) in FIG. 1C. The inventive device 100 according to FIG. 1C uses a look-up table 192 to obtain the average opinion score of voice based on the packet error rate (PEP), and uses a look-up table 194 to obtain based on the random packet loss rate (%) and data rate (kbps) The average opinion score of FTP and the video lookup table 196 are used to obtain the average opinion score of the video based on the packet error rate (PEP). The look-up tables 192, 194, and 196, or more generally, the relationship between the average opinion score and the transmission characteristics or parameters will be explained in more detail below.

[0081] Based on the scores obtained for each voice transmission, each FTP transmission, and each video transmission, the device for determining of the inventive device is each of voice transmission, FTP transmission, and video transmission, and determines the respective Transmission strategy 132, and publish the decision, that is, the optimal transmission strategy, to each layer, that is, the application layer and the wireless link layer in FIG. 1C.

[0082] The traditional method of determining voice quality is to use a human listener panel to perform subjective tests. The results of these tests are averaged to give an average opinion score (MOS), but these tests are expensive and impractical for online voice quality evaluation. To this end, the ITU has standardized a new model, Perceptual Speech Quality Evaluation (PESQ), which is an algorithm that predicts the quality score that will be given in a typical subjective test with high correlation. This is done by conducting an intrusive test and processing the test signal via PESQ.

[0083] PESQ measures the one-way voice quality: a signal is injected into the system under test, and the degraded output is compared with the input (reference) signal by PESQ. The mapping between average opinion score (MOS) and user satisfaction is shown in Figure ID.

CN 101026552 Β

[0084] The PESQ algorithm is computationally too expensive to be used in real-time scenarios. In order to solve this problem, a model that uses a small number of parameters to estimate the average opinion score (MOS) is proposed. These parameters (packet error rate and available bit rate) are easy to calculate. The available bit rate determines the speech codec that can be used. The experimental curves of the average opinion score (MOS) estimation as a function of the packet error rate of different speech codecs are shown in FIG. 2. Use a large number of voice samples and the average value of the channel realization (packet loss mode) to draw the curve. These curves can be stored in the base station for each supported codec. If you need to decode from an unsupported codec, these curves can be signaled to the base station as auxiliary information.

[0085] In order to estimate the satisfaction of FTP users, the paper used by A. Saliba, M. Beresford, M. Ivanovich and P. Fitzpatrick ^Measuring Quality of Service in an Experimental Wireless Data Network Australian Telecommunication Networks and Applications Conference, Melbourne, Australia, December 2003 The logarithmic MOS-throughput relationship introduced in the month (below [SalOl]). Assume that each user has subscribed to a given data rate, and his satisfaction is characterized by the actual rate he receives. Estimate the average opinion score (M0S) based on the current rate and packet loss rate provided to the user by the system:

[0086] MOS = a*logio[b*R*(l-PEP)] (3)

[0087] If the user has subscribed to the bandwidth R and received the bandwidth R, then in the case of no packet loss, its satisfaction based on the average opinion score (MOS) scale should be the maximum value, that is, 4.5. On the other hand, define the minimum bandwidth that can be provided to users and assign a mean opinion score (MOS) value of 1. By using parameters a and b, fit a logarithmic curve for the estimated mean opinion score (MOS). By changing the packet error rate (PEP), the model generates the average opinion score (MOS) estimation surface of Fig. 3 for each user at a reservation rate of, for example, 192kbps.

[0088] For example, the fitting of parameters a and b is completed in such a way that for the 192kbps ftp service as shown in FIG. 3, when the user receives the reserved 192kbps bandwidth without packet loss, the maximum value is obtained MOS 4.5, and when the user's actual bandwidth is Okbsp, the actual parameters of the minimum MOS Ε are selected as follows: a = 2. 6902 and b = 0.2452/kbpSo

[0089] In order to support video conferencing or real-time video in wireless multimedia networks, a simple model for evaluating the quality of video materials is introduced. Assuming that all information about the distortion caused by time slice loss is known, and the peak signal-to-noise ratio (PSNR) for different time slice loss percentages is evaluated. The model is constructed for Foreman video sequences, which are used as benchmarks Standard video sequence for benchmarking, but the model can be easily extended to different videos.

[0090] Use H.264JM & 4 codec for encoding and decoding. The encoder is set to encode the first frame as I-frames, and all subsequent frames are encoded as P-frames. It is assumed that there are 9 time slices per frame, and in each frame, the macroblocks of a single time slice are intra-coded (Figure 4A).

[0091] This leads to a higher bit rate, but also gives a higher resilience against packet loss (time slice). If the time slice is lost, after a maximum of 9 frames, the impact of the loss is cleared. In our experiment, for 0 percent packet loss, the average PSNR obtained for all 400 frames is 35. 30 dB.

[0092] FIG. 4B presents the relationship between the decoded average PSNR and the user satisfaction measured using the metric average opinion score (MOS). FIG. 4C shows the average opinion score (MOS) in the case of packet loss on the entire wireless channel. Each time slice is encapsulated into a packet. Use random time slice loss mode to simulate 1000 times for each% time slice loss. The average value of the decoded PSNR is calculated for all decoded frames. Use time slice or frame concealment and measure the expected peak signal-to-noise ratio (PSNR) and average opinion score (MOS).

CN 101026552 Β

[0093] For application-driven cross-layer optimization, three groups of users are defined: U-request voice service, V-file download, and W-video conference. Depending on the service, mobile users need different resources on the wireless channel. This depends on the channel code rate and transmission rate set available to the user. This is also called a transmission strategy. For example, you can use different voice codecs to serve users who request voice services (G.711, Speex, iLBC (Internet low bit rate codecs), or G. 723. 1.B in our example), and In our example, the data can be coded with different channel coding rates of 1/2, 1/3 or 1/4. Each transmission strategy gives users a different quality of service and requires a different amount of channel resources.

[0094] Create a transmission strategy set for each service. A is a collection of transmission strategies for voice services, Τ_νIs a collection of transmission strategies for file download services, and T_wIt is a collection of transmission strategies for video services.

[0095] The purpose of the optimization and the maximization of the average opinion score is to achieve maximum user satisfaction and fairness between users. For each user, according to the service, define the decision variable for each transmission strategy-whether to use the given transmission strategy to provide service to the user. Therefore, these decision variables are of Boolean type, that is, whether the user uses this strategy to send his information. For voice users, the decision variable is Uj, where "i" represents the i-th user, and "j" represents the transmission strategy that can be used for voice users. The next step is to associate the defined desired user QoS with the average opinion score (MOS).

[0096] Each user in a wireless network has a different location and mobility, which results in a varying receiver SNR. Based on the receiver SNR, for different modulation schemes (BPSK-Binary Phase Shift Keying and QPSK-Phase Shift Keying) and different channel coding rates, that is, for different modulation schemes (BPSK-Binary Phase Shift Keying and QPSK-Phase Shift Keying) -Elliott model, Technical Report TUM-LNS-TR-03-05, Institute for Circuit Theory and Signal Processing, Munich University of Technology, May 2003 (hereinafter referred to as [IvrOl]) in all transmission strategies, you can get the Estimated packet error rate. Generate channel realizations, and estimate the packet error rate (PEP) for all transmission strategies that are given the received specific SNR.

[0097] The objective function for multi-user multi-application cross-layer optimization is defined in formula (4). The sum of QoS (MOS) perceived by each user in the multimedia wireless network must be maximized. The parameter λ is used to give high priority to a given user, and its value is chosen by the network operator.

[0098] MaximizeXS/IMOSg] + Σ+Σ Σ heart w^MOS order (4) ieU ratio ieV j^T_v f recognize jet_w

<td>[0099]</td><td>The conditions are:</td><td></td><td></td>

<td>[0100]</td><td></td><td>Vzet/</td><td>(5)</td>

<td>[0101]</td><td>X x^ν·7 ^{= lj}</td><td>v/er</td><td>(6)</td>

<td>[0102]</td><td>ΣΧ =i &gt;j arc</td><td></td><td>(7)</td>

[0103]

ΣΣ^-ΣΣ^·+ΣΣ^ s total symbol rate (8) ieC j arc/eZ j^Ty

[0104] In the illustrated example, each user must be associated with a given transmission rate, channel code rate, and modulation scheme. The decision variables U, V" and W are of Boolean type, which means that the sum of all decision variables used for a user must be equal to 1, see formulas (5) to (7). The total available symbol rate of all users is limited to Less than the total symbol rate of the system. Each transmission strategy has an associated symbol rate and the sum of all individual symbol rates must be less than or equal to the total symbol rate, see formula (8) ο

CN 101026552 Β

[0105] In a preferred embodiment, the parameter λ is inserted to ensure fair allocation of resources. The optimizer tries to find a resource allocation that maximizes user satisfaction based on the mean opinion score (MOS), and this is usually the goal of every network operator. In this case, there is a possibility that even if the system performance is maximized, a given user cannot be satisfied. This may be caused by low receiver SNR, and the optimizer can decide to allocate resources to other users. This contradicts the fairness that should be provided to users independently of their location. In order to solve this problem, a scaling coefficient λ based on the history of QoS estimated by the user is selected. In each rate allocation process, find the user with the largest average value of the estimated QoS for the previous multiple steps, assuming that the largest average value is in the rate allocation step "j", and K users are in the system, a single The user's maximum perceived QoS value is obtained by the following formula:

[0106]

[0107]

[0108]

[0109]

[0110]

[0111]

MaxMOSj = max(£ MOS_U; £ M0S_2i MOS_Ki) (9)

1=1 i=li=l or introduce k as user or transmission subscript:

jj-1

MaxM0S_} = -- max(workM0S_Xi M0S_ki MOSQ (9) j ~ 1 &lt;=1 z=l/=1 Use the following formula to calculate λ for each user or transmission:

MaxMOS _fFu=---ZMOSh k = 1... K(10) The user with the largest perceived QoS has a scale factor of 1. Other users have scale coefficients in the range [1,4.5], because the denominator is also limited to the interval: 1; MaxMOSj]. This is important to maintain the stability of the optimization algorithm. Therefore, these λ values are the average opinion score (MOS) estimated by each transmission strategy scale, and maximize the sum of the average opinion score (MOS) of all users. The optimizer tries to assign a transmission strategy with a high estimated average opinion score (MOS) to users with a higher λ. This gives higher priority to users who have received lower QoS until the optimized moment.

[0112] The shared network performance metric is the throughput of the system. Traditionally, the goal of network operators is to maximize network throughput. Using throughput, at time j, the effective rate (goodput) of a given user i is 0_υ.Can be:

[0113] Gi"·=R"*(1-PEP) (11)

[0114] where R" is the actual transmission rate. The objective function used in the optimization model is to maximize the sum of the rates allocated to all users in the system, and is given by equation (12). Here, the optimizer does not know the user Perceived quality. Assuming that if the user receives a higher data rate, then he also has a higher QoS.

[0115] For throughput maximization, the same set of decision variables as in formulas (4)-(8) is used. The difference is the lack of scale parameter λ. There is no need to scale the allocated transmission rate, because the transmission rate required by different applications is incomparable.

<td rowspan="2">[0116]</td><td colspan="2">Maximize^ &amp;Yu+Working Dagger+Begging ZXq. (12)</td>

<td>/et/ jeTy</td><td>i^V j&amp;T_v ieW JgT_w</td>

<td>[0117]</td><td>The conditions are:</td><td></td>

<td>[0118]</td><td>Y knows=1, Vz Gt/ are</td><td>(13)</td>

<td>[0119]</td><td></td><td>(14)</td>

<td>[0120]</td><td>Busy 1, Viewj^W</td><td>(15)</td>

CN 101026552 Β

[0121]

Σ Σ ) + Σ Σ Ah + Σ Σ and ν total symbol rate (μ) ieU JeTy JeTy

[0122] FIG. 5A shows a diagram of a simulation architecture that compares the performance and user perception quality of an exemplary embodiment using the inventive method with the throughput maximization method. The simulation is performed using the following parameters: four voice users , Two male voice users Voicel and Voice2, and two female voice users Voice3 and Voice4.

[0123] The voice sample is 30 seconds long. The voice signal comes from a backbone network encoded at a rate of 64kbps using the G.711 voice codec. In the base station BS, according to the optimized output, the G.723.1 codec can be used to decode the signal to 6.4kbps, the iLBC codec to decode it to 15.2kbps, and the Speex to decode it to 24.6kbps, or It is sent at 64kbps without decoding.

[0124] Two users FTP1.FTP2 use FTP to order file downloads. Both of them subscribe to the service at the maximum transmission rate of 192kbps provided.

[0125] A user video requests a video conference. The video sequence used is Foreman encoded with H·264 encoder. The frame sequence is the appropriate format for real-time video jppp-.. -P_o

[0126] The total available system rate is constant, and three different conditions need to be checked: 500k symbols/sec, 700k symbols/sec, and 900k symbols/sec. The supported modulation schemes are DBPSK (differential BPSK) and DQPSK (differential QPSK). Support one-half, one-third and one-fourth channel coding rate.

[0127] Because of the user's mobility, the SNR received by the user for each optimization step is randomly drawn according to the unified release from a given interval_OThe system is active for 30 seconds, and assuming that the average channel characteristics remain constant for 1.2 seconds, this results in 25 optimized loops.

[0128] In order to obtain the relationship between SNR and PEP, a Rayleigh fading channel is simulated. For a specific combination of signal-to-noise ratio (SNR), modulation scheme (DBPSK or DQPSK) and channel coding rate (1/2, 1/3 or 1/4), the transmission of one million symbols on the channel is simulated . For this particular setting, calculate the residual bit error rate (BER) after receiving symbols at the receiver. Based on the bit error rate (BER), the packet error rate (PEP) is calculated using the application layer packet size. For simulation, 640 bits are used for G.711-encoded packets, 304 bits are used for iLBC-encoded packets, 496 bits are used for SPEEX packets, 192 bits are used for G.723.1 packets, and 640 bits are used for FTP packets. And 900 bits are used for video packets.

[0129] The scaling factor λ is calculated based on the expected MOS (not the actual MOS) in the previous optimization step. Therefore, it is assumed that there is no application layer quality feedback from the mobile terminal or user to the base station.

[0130] For voice users, the signal samples are divided into multiple 1.2 seconds, and each sample is coded using the voice codec given from the optimization algorithm. At the end of the optimization loop, these voice samples are collected into a single file, and the perceptual quality (MOS) is calculated by comparing the original signal and the distorted signal.

[0131] For video users, if the time slice is lost, the time slice is not written into the bitstream, which will tell the decoder to call the error concealment algorithm. Calculate the PSNR of each frame and the resulting average PSNR. The average PSNR is converted to an average opinion score (MOS) value using the relationship shown in FIG. 4B.

[0132] FIG. 5B shows a flowchart of an exemplary embodiment of the inventive method for the simulation architecture as shown in FIG. 5A. Figure 5B shows seven steps S510 to S570 repeated for the simulation.

[0133] In step S510, seven mobile terminals-each "user" has one terminal and executes one application-receive the first to third transmission types (application types: voice, FTP, and video). The first to seventh transmission data (application data: 4x voice, 2x FTP.lx video) associated with the seventh transmission (application: 4x voice, 2x voice, lx voice)

CN 101026552 Β

frequency).

[0134] In the subsequent step S520, for each single transmission, the mobile terminal independently obtains the transmission characteristics of the received transmission data: transmission rate, packet loss rate and/or signal-to-noise ratio. Based on this, the desired MOS can be determined for different applications.

[0135] In step S530, based on the transmission characteristics of each transmission as described in [Ivr01], a packet error rate (PEP) is derived. This step can be performed by a mobile terminal or a base station.

[0136] In step S540, the user-perceived quality score (average opinion score MOSQ) is obtained or derived based on the transmission characteristics of the specific transmission mode, which means that the transmission specific mode is: for voice (or more) based on the packet error rate according to FIG. 2 Generally, according to the specific method of voice represented by the lookup table 192 in FIG. 1C; for the specific method of FTP based on the packet error rate or packet loss and data rate according to the lookup table of 194 in FIG. 3 or FIG. 1C; A video based on the packet error rate defined according to the lookup table of FIG. 4B or according to 196 of FIG. 1C.

[0137] In step S550, the base station BS calculates the scale or fairness factor kJ for each transmission based on the historical score (MOSQ) according to formulas (9) and (10) (or the formula (4)^Xui> ^λνί> ^Xwi)°

[0138] In step S560, the base station BS maximizes the total expected score (E[MOS") to determine the optimal transmission scenario, that is, for each transmission, for voice, for FTP, and for video The optimal combination of each transmission strategy Uij, Vij and Wjj.

[0139] In step S570, it is determined that u",_ViJAnd w" represents the optimal transmission strategy, as well as the various transmission parameters (such as source codec, channel codec, modulation scheme) used in the application layer and the wireless link layer, and the various data rates available for each transmission , The base station sends transmission data (4x voice, 2x FTP, lx video) in the subsequent transmission interval based on the determined transmission strategy (Uij, Vij, wQ).

[0140] After step S570, perform step S510 again based on the newly transmitted data

[0141] In the following, a comparison is made between the two investigated optimization methods. The architecture described in the previous section was used, and each simulation was run 600 times.

[0142] FIG. 6 shows that between the average opinion score (MOS) maximization, that is, the embodiment of the inventive method according to FIG. 5A, and the throughput maximization rate allocation scheme, the voice users Voicel to Voice4 The improvement of voice user satisfaction. At a total system rate of 500k symbols/sec, the average gain in terms of average opinion score (MOS) is 0.85. At 700k symbols/sec, the gain is still significant-0.6, and for 900k symbols/sec, it is approximately 0.4. The Mean Opinion Score (MOS) maximization scheme results in a small improvement in the increase in the available transmission rate. This means that in the absence of resources, it provides users with good quality.

[0143] FIG. 7 shows the gains of FTP users: FTP1, FTP2. The Mean Opinion Score (MOS) maximization method is again superior to the throughput maximization method. Here the gain is lower, but still significant. For 500k symbols/second, the average gain is 0.7MOS, for 700k symbols/second it is 0.45, and for 900k symbols/second it is 0.3.

[0144] FIG. 8 shows the improvement of the video conference quality of the video user Video. The gain in terms of MOS is similar to the figure

The voice user Voicel to Voice4 in 6, and as the available transmission rate increases, the gain decreases.

[0145] For all the cases shown in FIGS. 6 to 8, MOS maximization has the advantage of providing users with a lower spread of QoS. For example, if you consider Figure 6 for a situation where the total system symbol rate is 500k symbols/second, the resulting MOS in 90% of the throughput maximization situation varies between 2 and 3.5, that is, 1.5 MOS Widen. On the other hand, maximizing M0S causes M0S to vary between 3.4 and 4.1, that is, only a spread of 0.7M0S.

[0146] In FIGS. 9 to 11, voice users Voicel to Voice4 are referred to as voice users 1 to 4, FTP users FTP1 and FTP2 are referred to as FTP users 1, FTP users 2, and Video users are referred to as Video users 1.

CN 101026552 Β

[0147] Figures 9-11 show the gain of each user in the system. As the difference between the MOS calculated using the MOS maximization and the throughput maximization, a curve is generated. Starting with a system symbol rate of 500k symbols/sec (Figure 9), in 50% of the simulation, the average gain for all users is 0.8. The exceptions are the video conference user Video, which has an even higher MOS gain, and the FTP user 2, which has a lower gain. In a 700k symbol/second system (Figure 10), there are cases (1% for users with video conferencing), where maximization of throughput gives better results for a given user. When two users (the user with a video conference and the fourth voice user) have better performance (10% of cases) with maximum throughput, this is at a system symbol rate of 900k symbols/sec (Figure 11) The situation is even more pronounced. The users mentioned are those that have the best channel with respect to the received SNR. In the case of MOS maximization, the optimizer takes resources from it to increase the average opinion score (MOS) of users with poor channels.

[0148] On the right side of the flowchart in FIG. 5B, an exemplary information flow between entities performing tasks is shown, where MT (mobile terminal) represents a mobile terminal or user (such as Voice 1 to Voice 1 in FIG. 5A). Voice4, FTP1, FTP2, video), and BS (base station) represents an entity like the base station responsible for resource allocation (such as the base station in Figure 5A). [0149] The solid arrows between entities describe the first downlink scenario, where the mobile terminal MT sends the transmission characteristics or the obtained packet error rate to the base station BS, that is, steps S510 to S530, or S510 to S520 are performed. Accordingly, the base station BS performs steps S530 to S570 or S540 to S570.

[0150] The dotted line between the entities describes the second downlink scenario, where the mobile terminal MT sends the score to the base station BS, that is, steps S510 to 540 are performed. According to this, the base station BS only performs steps S540 to S570.

[0151] The dotted arrow describes the uplink scenario. After step S560, an additional step is required. In this step, after the optimal transmission strategy has been determined, the base station BS sends each transmission strategy u", ν", w" to each mobile terminal MT, and then each mobile terminal MT The terminal sends transmission data based on this transmission strategy in the subsequent transmission interval.

[0152] Typically, the preferred embodiments are implemented so that mobile terminals only perform minimal processing, because they typically have only limited processing capabilities compared to base stations, and parts or steps that require strong processing capabilities are performed at the base station. Thus, in a preferred scenario, the mobile terminal MT will only perform steps S510 to S520, send the transmission characteristics to the base station BS, and the base station BS will perform the remaining steps S530 to S570.

[0153] Depending on the application or transmission type, the mobile terminal MT can even perform further steps for, for example, each transmission, for example, S530 or S540, to determine which entity performs which steps.

[0154] Although FIG. 5A only shows a downlink scenario, as mentioned above, the present invention is not limited to a downlink scenario, but can also be used in an uplink scenario and a mixed uplink/downlink scenario. Link scenarios are used for any number of users, applications, and application types, and are also used for users who perform more than one application at the same time.

[0155] In another embodiment of the present invention, a priority coefficient Wk is used. The priority coefficient Wk represents the relative importance of the user, for example, as determined by the service agreement between the user and the service provider, where k=1. .K is one of K users. In another embodiment, the priority coefficient will be not only user-specific but also application-specific, that is, the service agreement not only defines a general application-independent priority coefficient for the user, but also applies to each application. Define a specific priority coefficient. The user can subscribe to the specific service level of the service, and based on this, get, for example, voice Wui, FTP_viAnd the application priority coefficient of video w". The priority coefficient can be used instead of fairness or scale factor, or used in combination with scale factor. Formula (17) shows the optimization function with priority coefficient:

[0156] Maximize" iwU JeTu Heart j^T_v Chemical0 j^T_w

[0157] As an alternative to the scenario described in FIG. 5A, the inventive device for determining the transmission strategy can be used in any other

CN 101026552 Β

It is usefully implemented in a device where any other device is responsible or allowed to be responsible for allocating network or radio resources.

[0158] Although the foregoing description focuses on parameters from the application layer and the wireless link layer, as shown in FIG. 1C, alternative embodiments of the present invention may also include parameters from other layers (for example, the transport layer or the network layer). .

[0159] In summary, the present invention provides a device and method for determining a transmission strategy, and a system that allows optimization of wireless network resource allocation across multiple types of applications. In a preferred embodiment, the present invention proposes an optimization scheme based on, for example, the user-perceived quality score and the average opinion score (MOS) of a unified or shared measure. The average opinion score quantifies users satisfaction with service delivery. The present invention can be used in any system that handles service delivery on the entire mobile communication network.

[0160] The present invention will be beneficial to increase the network capacity, that is, provide services to a large number of users at the same time, and will improve the user-perceived quality of service (QoS).

[0161] Using Mean Opinion Score (MOS) as an optimization parameter is beneficial in different ways. First, this enables simple and direct fairness measures to be given. Secondly, because it is based on the same scale of application layer performance, this allows not only the diversity of the physical layer to be used, but also the diversity of the application layer. In addition, this may provide a highly flexible framework for cross-layer optimization, such as adapting applications to transmission, network, data link and physical layer characteristics (inversion methods), and adapting the physical, data link and network layers to application needs (Sequential method). In particular, the present invention is very beneficial to network operators because it allows improving user-perceived QoS and increasing network capacity by maximizing the number of users that can be served at the same time.

[0162] According to the specific implementation needs of the inventive method, the inventive method can be implemented in hardware or software. This implementation can be performed using digital storage media, especially optical discs, DVDs or CDs on which electronically readable control signals are stored, which cooperate with a programmable computer system to perform the method of the present invention. Therefore, in summary, the present invention is a computer program product having a program code stored on a machine-readable carrier, and when the computer program product runs on a computer, the program code is operable to perform the method of the present invention. In other words, the method of the present invention is therefore a computer program with a program code, and when the computer program runs on a computer, the program code executes at least one method of the present invention.

CN 101026552 Β