WO2002089513A1

WO2002089513A1 - Quality of service state predictor for advanced mobile devices

Info

Publication number: WO2002089513A1
Application number: PCT/EP2001/004655
Authority: WO
Inventors: Vahid Mirbaha; Ramin Mirbaha
Original assignee: Fg Microtec Gmbh
Priority date: 2001-04-25
Filing date: 2001-04-25
Publication date: 2002-11-07
Also published as: WO2002089516A1

Abstract

A method for predictive computing of variable mobile link parameters per session and state in a future time interval within Radio Access Networks (RAN). The system narrows its prediction errors as time progresses. An ideal control value is also estimated, so that mobile QoS applications can determine their intervention point while coexisting with link layer resource management mechanisms. The system is provided with certain measurement elements from the RAN. With the information contained in these elements, the system estimates the Received Signal Strength (RSSI), Received Wideband Power (RSCP), Signal Interference Ratio (SIR), Bit Error Rate (BER), the transmission delay per frame (delay), the variation between delay measurements throughout a certain number of measurement time frames and the mean bit throughput rate (Chip Rate) for a future time interval. The system also calculates the optimal power control (gain) for a defined value target of a given link parameter. The estimates are passed on to other Systems for further processing.

Description

QUALITY OF SERVICE STATE PREDICTOR FOR ADVANCED MOBILE DEVICES

Description of invention

Background of invention

Technical Field:

The present invention relates in general to the prediction of changes in a mobile communications environment. In particular to a method of predicting mobile link characteristics while at least one party is in motion. Still more particularly, the present invention relates to a method of predicting Link Quality Parameters in 2.5 and 3G mobile access networks, considering lower layer corrective mechanisms such as power control, to aid QoS systems and Applications in their quality management process .

Description of the Related Art :

The standardization of wireless systems beyond the current se- cond-generation is rapidly progressing in all major economic regions of the world. These systems are known under names such as IMT-2000 (ITU), UMTS (ETSI 3GPP) , EDGE and ANSI 3GPP2. While current systems such as GSM, PDC, ISI36 and IS-95 have been used for circuit oriented voice telephony, the newer gen- eration of mobile access networks also known as 2.5 and 3G will off more bandwidth and services. The main application for these services will be wireless packet transfer. The transport of IP (Internet Protocol) packets over the air interface not only extends the reach of the internet to the mobile user in a known and trusted fashion, it also opens the opportunity to migrate all of the communication to a packet switched environment. By gradually eliminating the need to establish separate logical circuits between the end device and the next mobile network node, the scarce radio resources can be put to work in a more efficient manner. This will lead to lower Network Oper- ati g Expenses (NETEX) and in turn to more attractive subscription or transaction models.

Using IP as the transport mechanism for mobile radio networks also has its challenges. Typical services with real-time re- quirements are packetized voice and video, as well as delay sensitive applications such as traffic signaling, remote sen- soring and interactive web applications. The challenge here is to provide acceptable quality while maintaining spectrum efficiency. "Acceptable quality" is what the human user preceives to be "good" . Voice applications such as voice telephony have been in use for a long period of time and certain delay, jitter, and loss boundaries are now known to be "good" . Any conversation with a one way delay of more than 150ms or 12-15% packet loss or more than 10 ms jitter is perceived to be de- graded or unusable. In particular, the areas of concern are: Spectrum efficiency Low latency Data integrity Sufficient bit rate

While spectrum efficiency is being addressed by robust compression schemes both for the payload and the packet header, the current invention supports and enhances existing schemes to achieve acceptable values for the latter 3 areas .

Some factors that have an influence on the link quality are: Voice activity: drives the codec mode and bit rate. Some codecs such as the AMR codec include voice activity detection (VAD) and generation of comfort noise (CN) parameters during silence periods. Hence, the codecs can reduce the number of transmitted bits and packets during silence periods to a minimum. The operation to send CN parameters at regular intervals during silence periods is usually called discontinuous transmission (DTX) or source controlled rate (SCR) operation. Loading: this is the effect of neighbor cells in different load states. A base station serving more than 60 subscribers in a Rural Urban area will transmit at high power levels, influencing the link quality in adjacent cells. Sectorization: In order to serve more subscribers, cells may be sectorized. This involves more hand offs called "softer" hand off .

Multipath fading: occurs as signals bounce of objects, arriving both directly and indirectly. This effect influences the delay boundries as signals arrive at different phases. The effect is a volatile BER due to varying SIR.

Power control mode: depending on the mode employed (open or closed loop) interference may occur with neighboring UEs . Cell and RAT hand offs: also called soft and hard hand offs influence the transport context and indirectly the subjective quality perception. In cases where the Radio Access Technology (RAT) is changed (e.g. from an UMTS access network to a GSM network), hard hand offs may cause large transmission delays. Terrain: whether the surrounding is an open plain or a moun- tainous region has an affect on the various propagation models. While generally radio propagation delays such as the multipath rayleigh fading have been regarded as quality degrading, this is to a certain extent different for CDMA based networks. In these cases, the effect of multipath fading can both degrade or upgrade the Signal to Interference (SIR) value. Depending on the dimension und duration of such effects, the user of the current invention will benefit from accurate SIR predictions . Radio Coverage: Obviously, the degree of cell coverage, espe- cially in sparsely populated regions, is one of the main contributors to link quality.

Velocity: most of the factors described above have a direct relationship with the speed with which the UE travels. Most notably, the hand off procedures and power control mechanisms are directly influenced by the speed and direction of the UE . Existing QoS Schemes address changes in the link quality through various methods . Most known systems attempt to adapt to the changing quality environment induced through mobility. The current invention offers an enhancement to these systems by providing future measurements, target measurements and the related control values that have not been part of the art previously.

Detailed description of invention

The present invention addresses the need to know quality vari- ations in the radio access network in advance. The quality variation arises through the motion of the mobile user, the engagement of resources, as well as sporadic disturbing factors relevant the radio access network. Quality state changes generally effect packetized realtime applications adversely. The nature of the fast changing link state within a wireless mobile environment requires anticipatory and pre-emptive measure to contravene the effects. Thus, providing link state information and control input values in advance is a valuable support for any QoS management system. The provision of such support is made possible by the present invention.

Broadly speaking, the estimations are achieved by observation and prediction. The system does not require specific knowledge of a relationship model to initiate the process.

The following steps describe the method of observing and computing ideal values for QoS Service Levels: 1. ^' Observe and record the following measurements. It is assumed that the following measurements are provided to the system: a. Received Signal Code Power (RSCP, [4,5]) b. Signal to Interference Ratio (SIR, [4,5]) c. Received Signal Strength Indicator (RSSI or wideband received power, [4,5]) d. Traffic Volume Measurement [1] e. one way transmission delay ([1, 6, ]) f. Block size [2] g. Block Error Rate [1] h. UE position, direction (bearing), altitude and velocity [7, 8]

2. Construct covariance matrices for the sum of all changing vectors

3. Use the known interdependencies to estimate future values for certain vectors, while minimizing the prediction error 4. Given optimal target vector trajectories (QoS profile), determine future control (s) values required, in order to a- chieve the desired trajectory. As an example: if the agreed

Service Level for the current QoS Profile includes a Bit Error Rate (BER) of 10-4, how much power gain is needed to equalize the effect of the Doppler shift, path and multipath fading, considering the relationship between motion, direction, Net- work coverage etc... An external system provided with this information can make an educated decision on the type of response it will make before limitations become obvious. Such responses could be in the change of coding rate or compression ratio. Other applications can us this information to change the protection method (Unequal Error Protection) or fine tune this to meet the predicted change.

Each measurement value has its own vector in time and its own trajectory in the future. This is also true for modulated sig- nals. In a real environment however, the number of influencing factors, dependencies and unknowns are often too large to allow a relational expression between measured values. Relationships are often instantaneous and only reveal a predictable pattern when covariances are analyzed over time. To this extent, the current invention introduces the following general model :

The main goal is to predict the future of a process given its past and covariance structure (from previous observations) . More precisely, the system is given a measurement (RRC measurement outputs of a given UE in a spatial domain)

y_t-ι* Yt - 2 , .... Yt - i presenting past inputs

presenting future outputs

wherby y can be any meaurements described in claims 2, 3 and

4.

More specifically, the goal is to predict the future link states described by link equality attributes (y**., ..., y_n) under influence or control ( u_{l f} ... , u_n)

The past and the future vectors of each QoS indicator is of interest. Let the covariance matrices

Lpp, ff, Lpf, ZJ (2)

be determined from previous observations. It is possible to determine the future prediction vector

_t= P (P_t ! ∑) = prediction of the future f_t

which minimizes the expected prediction error f_t - f_t . Here the metric to measure the error is chosen in a stochastic sense, e.g.

Where E{ } is the expectation operator and | | | |_s is a suitably chosen weighted norm. The norm is a function that measures the size of the vector, so, for instance, | \ f_t~f_t \ is under- stood to be the size of the prediction error f_t~f_t -

Up to this point it can be said that given observations describing the state change pertained to a known number of qual- ity indicators and observing the corresponding control inputs such as dynamic power management, it is possible to use the past (written as p_t in relation (1)) to compute the prediction

Λ vector f_t of the future values such as error rate and signal strength, where the prediction is optimal in the sense of re- lation (3) , that is the expected prediction error of f_t is minimal over all possible predictions. The system allows for any number of control inputs (these could be power, CPU, buffer or memory related) . The presented algorithms are based on well known covariance techniques. The current invention uses the canonical variate analysis [12] . The past p_t and the future f_t are considered random vectors with an observable covariance structure. The computational part is based on the Generalized Singular Value Decomposition (GVSD) , which is also used for optimal reduction of the dimension of the problem. The invention makes use of different algorithms for computing the GVSD values. A detailed description can be found in [13] . The choice depends on the resources available to the system embodying the current invention.

We now have an optimal prediction of RF and link quality values for a future time interval which is n- -r*_!.

These predictions include the behavioral influence of control inputs such as dynamic power control. However, in certain circumstances, the corrective influence of a control mechanism (i.e. fading compensation) may either have natural or given limits. This is the case for equalizing the signal fade through increase of transmit power until limits set either by the operator or legal bodies are reached. In other instances the gain in transmit power may interfere with other subscribers or more commonly induce self interference which will in turn increase the Bit Error Rate. Although in most cases cell hand off procedures relax this situation, these procedures induce quality degradations of their own. Other control mechanisms such as Forward Error Correction (FEC) have limitations specific to the employed method. In this case it is the addi- tional bandwidth required to transport the FEC packets.

Considering these limitations of control mechanisms, the question that arises is: at which point will a corrective measure either produce incremental results below minimum expectancy or generate side effects such as interference or delay, which are undesirable .

In order to answer the question the system determines the mi- nimal future control required to control the trajectory of the future outputs. Here we deal with input (y) , control (u) and the desired trajectory (y) .

Functionally speaking, it is desired that, in the future, the variable y behaves as prescribed by y, e.g. it is desired that the one way transmission delay between RLC frames remains below a certain value. Having previously analyzed and determined the covariational relationship between signal strength and transmission delay for this specific session (including the covariations of BER, velocity and direction) the system now determines the required future control for such an equilibrium.

Let u_t ⁺ and y_t ⁺ denote future control and future desired tra- jectory, respectively, and let f_t ⁺ be the future trajectory due to the future control. The goal is now to minimize

^E { | | yt⁺-Λ⁺ M M ² _S } + l l -V M ² ,

with suitably chosen norms. More precisely, the problem is how to choose minimal future control to get as close as possible to the desired trajectory. The algorithm used in this case are described in [15] .

An Example:

Let us assume the described method is put to use with the following value types: RSSI, SIR, BER, Rx-Tx delay After an observation period covering ideally at least 80 samples, and applying the described methods, we arrive at a RSSI_t+1 and transmit power gain (power control) PC_t+1. These and other predictions are provided to external applications using an output frame described in figure 4. In case of an application using the AMR WB codec, the PC_t+1 value would be used to determine the point in time for intervention. At such time , the external QoS Management application may decide to replace the current RSSI value contained within the current AMR frame with the predicted RSSI_t+1 supplied by the current invention.

The codec bit rate would be adapted to a link state just about to occur . Of course, this would only apply to the receive side of the codec signaling a mode change to the encoding peer. However, applied correctly, the pre-emptive mode change would lower the amount of residual bit error per block of information relevant to the codec.

Brief description of the drawings

Figure 1, is a table of QoS quality parameters widely used in Radio Access Networks.

Figure 2, is a table showing control inputs (power control) and the respective tolerance ranges

Figure 3, generic view of codec mode changes relative to channel power.

Figure 4 is a diagram illustrating the Information Element (IE) containing the output of the state predictor

Figure 5 is a block diagram showing the steps involved in mak- ing link state predictions Figure 6 is a block diagram of QoS Management System advantageously using the current invention.

Figure 7, is a block diagram depicting the layer positioning of the current invention relative to the 3GPP model.

Figure 8, is a theoretical output graph employing the algorithms described in this invention.

Detailed description of drawings

Referring to figure 1, listing the QoS classes defined by the 3GPP working groups, for which the Universal Terrestrial Radio Access Network (UTRAN) has been designed. Each class shows a large range of acceptable values. Of particular interest for real-time applications is the residual BER, which is defined as the error rate that is not detected or corrected by lower layers and actually effects the application. With a residual BER of 10^"4, a frame length of 1500 bytes and a packet length of 256 bytes, the application could be confronted with as much as 15-20% packet loss. The current invention is designed to assist QoS management systems in dealing with these type of situations as early as possible or useful.

Referring to figure 2, a reference table consisting of the power control level 101, which denotes the control state the power management unit can take. The nominal output power |20_ is the power the UE must transmit when commanded by the respective power control level. Although the power control may exceed this in some cases, the maximum specified and permitted power output is generally not higher than 33dBm.

The tolerance range BC shows when a power control level is changed and so a new control command issued. The reader can observe the tolerance range is lower at higher output numbers. This is related to the rise of interference, which is directly related to the transmitted power. Hence, in a scenario in which the UE travels away from the serving base station, a na- tural limit is set to the point in time when a drifting cell must have been selected for hand off. Similarly,' in closed environments such as cars and trains, power levels at 6 and below may cause interference with other UEs . In such cases a hand off may not occur, as the hand off thresholds may not ha- ve been reached. However, the increased Bit Error Rate would either require a pre-mature hand off or a lower transfer rate. This is a common scenario when traveling with public transportation means in rural areas.

Referring to figure 3, a diagram showing thresholds for codec mode changes. Two AMR codecs with different capabilities are used. ID and 1U show thresholds levels for a 3 mode codec (5,9; 7,95 and 12,2 Kbps) where D denotes the downlink connection a U the uplink connection. Similarly, the curves 2D and 2U show applicable thresholds for mode changes relative to a 4 mode AMR codec. The current invention makes use of the described model, in order to support and invoke a mode change immediately prior to the link state change.

Referring to figure 4, an overview of the frame structure used to present the results of the current invention to external applications. The frame is denoted Information Element (IE) for compliance reasons to 3GPP specifications. It is organized octetwise .

Type Comment Octetl

Bits 1-2: This contains the frame type. Frame types are defined as follows: 1 3GPP release 99 conforms with 25.215 sub- clause 5.1 and 25.225 sub-clause 5.1 2-0 reserved Bits 3-8: reserved

Octet2

Bits 1-4: reserved

Bits 5-8: size of estimation interval in ms . The window size depends on the size of the prediction error from the past n predictions while applying a progressive weighting mechanism in order to weigh heavier on the most recent estimates. Future extensions of the current invention may also change the windows size based on factors such as terrain, speed or application specific requirements .

Octet3

Bits 1-2: Signal to noise ratio (SIR) as defined in 3GPP TS 25.331 v. 3.5.0 expressed in dBm. This value contains the estimation of SIR within the estimation window which is used by various applications such as the AMR codec as an indication of possible BER values. Bits 3-4: Received Signal Strength Indicator (RSSI) as defined in 3GPP TS 25.331 v. 3.5.0.

Bits 5-8: Received Signal Chip Power (RSCP) as defined in 3GPP TS 25.331 v. 3.5.0.

0ctet4

Bits 1-7: reserved

Bit 8: Bit Error Rate (BER) 1-9, where each integer represents n in lOE-n

0ctet5 Bits 1-4: Prediction Accuracy expressed in 3 digits and interpreted as percent, relative to the progressive weighted average of all estimates of t-1 compared to measurements of t-1 Bits 5-8: Transmission Delay as defined 3GPP TS 25.331 v.

3.5.0, expressed in ms .

Octet6

Bit 1: Control value type, integer, 1 bit. Definitions are:

1. Power control level

2. Class A protection length (UED/UEP) according to [16]

3. Codec Mode according to UE capability statement 3GPP 21.904 v. 3.3.0 and codec type

4. Multiple frame encapsulation according to [17]

5. Robust header Compression (ROHC) [10], mode

6. Robust header Compression (ROHC) [10], state

7-9. reserved Bits 2-4: Control value quantity depending on supplied type:

Type: 1 = power control level 2 = MSB Coding point in absolute bits, left to right

3 = Codec Mode depending on UE statement

4 = number of multiple frames in RTP 5 = ROHC mode type (001 through 009)

6 = ROHC state type (001 through 009) Bits 5-7: Profile number to be determined by the QoS Management System and current invention Bit 8: 0-9 reserved Referring to figure 5, a simple overview of the main processes of this invention. The measurement collection 10 is done via the C-SAP of the RRC UE control agent [1] .

Once the a minimum number of input values have been collected, their progression in time is observed 20. Not only the time based change per singular value (such as BER) is of interest but also variations between these values, that show a correlation. The invention does not require an initial model of the relationships between the measured vectors. With other words, it is not a prerequisite to provide a relevant propagation model to initiate the analysis. The reader will appreciate the "data in - model out" approach employed within the invention. Although the exemplary embodiment is based on a UMTS environment, it can be applied to any mobile system. This allows the addition of any new input type or control mechanism without the need for a reference model. Similarly, any desired quality value can be idealized, delivering the prediction capabilities to a large number of mobile environments. In order to analyze the instantaneous covariations among the input vectors (which progress continuously in time) a specific statistical method called CVA is employed. More specifically, within the CVA mechanism, a generalized singular value decomposition which transforms a basis set of input variables and future output variables to correlated random variables is em- ployed. The matrices are obtained via a singular value decomposition (SVD) of the cross-covariance matrix. The exact method is described in [15] . Provided sufficient number of observation samples, the operation provides an accurate expression of the momentary relationship and dependencies between all input variables. With other words:

Given speed, direction, RF values such as RSSI, RSCP and SIR, given FER, transmit delay, bit throughput and sufficient samples, the "predict change" 20 module will a) predict the how these values will change and, more importantly, b) express the interdependencies for a given time interval. We now know the state of the link quality in the future. To a system without an explicit statement of the desired QoS parameters (QoS Profile) this information is presented as an Information Element (IE) containing the estimates |40|. The IE is described in detail in explanations to figure 4. In a mobile system without corrective control, pure estimates of state changes are sufficient to determine adequate equalization measures. For example, in a mobile system without any power control, the future estimates of BER and delay can be used to adjust compression rate and frame lengths in advance.

Given a certain QoS profile, it is of interest to determine the ideal control required to satisfy the profile. For this purpose the profile is interpreted as a desired trajectory of the input values 50. If the system does not foresee corrective methods, the invention reverts back to the simple IE 40. However, in most current and planned mobile systems there is at least the element of power control present. In this case, the question that is of interest is: Will the corrective measure, inherent to mobile system, satisfy the QoS or Service Level desired by the application? If not, at which point in time should an additional QoS Management System (if present) or o- ther additional corrective measures be activated and to which extent?

The answer is influenced by the following factors: 1. The interdependencies of all time sensitive variables

2. The QoS Profile itself.

3. The effect of the corrective measure in question on the system and the prediction accuracy. 4. The complex effects of the plurality of corrective measures in the over all mobile system at that point in time. Module 60 in figure 9 addresses the first 3 areas. A simple description is provided in a previous section of this docu- ment. For a detailed explanation, the reader skilled in the art is advised to refer to [15] . The output of this function assumes that further analysis of the computed ideal future control is carried out by upper layers (QoS Management systems) with regards to Radio Access Technology (RAT) and the specified quality driven actions such as Radio Resource Management (RRM) strategies. The output of module 60 is the prediction frame 70 which is described earlier in this document (see description of figure 4) . This frame is identical to that of 40, except for the information contained in octet6.

Referring to figure 6, a conceivable QoS management model designed as an UE stand alone client incorporating an exemplary embodiment of the current invention denoted here as the "State Predictor" . The following description is concerned with the interaction and practical use of the current invention in such a model. The UE measurements 20 are collected and forwarded via an interface manager 50. After establishing the session, the measurements are forwarded to the state predictor 60. Estimates concerning the future state of the current link ex- pressed through variance estimates of the original input data are presented to the service level manager using the output frame format described in figure 4. The output frame also contains control level estimates such as Tx power gain. The latter is used by the SLA Manager to calculate the best point in time for each given action. Certain estimates contained in the output frame are forwarded to higher layer applications for further processing. The Service Level Manager will use the state predictions to set appropriate transport and compression protocol variables ahead of time 40 and 70.

Referring to figure 7, the position of the current invention within a layered protocol model . Measurements from the radio transmission layer 30 are provided through the interface control agent C-SAPI of the RRC UE agent or any other conforming uppler layer application , e.g. QoS Management System. Link state presdictions concerning packet transport quality 50 are provided to the packet transport layers, while in certain cases RF quality indications are forwarded via the QoS Management system to the codec 10 or the application level.

Referring to figure 8, a diagram generated using GNUPLOT. The graph shows generated signal and its prediction starting at mark 40 at the X axis. Here the mark 40 is the point where the future starts. The past is not shown (0..40) -- the past is used only to project into the future. The covariance matrices are computed in an off-line generation of samples. Here, of course, all data are experimental and serve the purpose of a model prediction output.

Quality State Predictor

Reference

[1] TS 25.331 "RRC Protocol Specification" [2] TS 25.322 "Radio Link Control (RLC) Protocol Specifica- tion"

[3] TS 25.321 "Medium Access Control (MAC) Protocol Speci- fication" [4] TS 25.215 "Physical layer - Measurements (FDD)" [5] TS 25.225 "Physical layer - Measurements (TDD)" [6] TS 25.932 "Access Stratum Delay Budget" [7] TS 25.305 "Stage 2 Functional Specification of UE Posi- tioning in UTRAN"

[8] TS 23.032 "Universal Geographical Area Description (GAD) "

[9] G. Golub, Ch. Van Loan: Matrix Computations, Johns Hopkins University Press, third edition, 1966 [10] Robust Header Compression (ROHC) : Framework and four profiles: RTP, UDP, ESP, and uncompressed <draft-ietf-rohc-rtp- 09. txt>

[11] EP 1 059 792 A2 : "Method and system for wireless QoS agent for All -IP network", Nortel Networks, 13.12.2000 [12] Larimore, W.E: (2000), "Identification of Colinear and

Cointegrated Multivariable Systems Using Canonical Variate A- nalysis, " in Preprints of Symposium on System identification 2000, held June 21-23, 2000, Santa Barbara, CA. [13] Golub, gene H. and Charles Van Loan, Matrix Computations, Third Edition, Johns Hopkins University Press, Baltimore, 1996 [14] TS 23.107*. "QoS Concept and Architecture"

[15] Wallace E. Larimore, Franklin T. Luk, "System Identification and control using SVD's on Systolic Arrays", SPIE Vol. 880 High Speed Computing (1988) QA 76.54 #54, 1988 [16] draft-ietf-avt-ulp-00.txt: "An RTP Payload Format for Generic FEC with Uneven Level Protection"

[17] draft-ietf-avt-rtp-amr-06.txt : "RTP payload format and file storage format for AMR and AMR-WB audio"

Claims

ClaimsWhat is claimed:

1. In a mobile communication system, a method of predicting and presenting the transport quality of packetized application data in a radio access environment consisting of the following steps : a. recording at least one RF metric in at least one time interval b. recording at least one motion metric in at least one time interval c. recording at least one packet transport metric in at least one time interval d. estimating at least one packet transport metric for at least one time interval in the future e. estimating at least one RF metric for at least one time interval in the future f. correcting estimation errors for at least one transport metric . g. correcting estimation errors for at least one RF met- ric. h. estimating at least one target value for at least one desired RF output. i. estimating at least one target value for at least one desired Transport metric output. j . determining minimal required control to achieve one of the desired outputs in h. k. determining minimal required control to achieve one of the desired outputs in i.

1. presenting the predicted output value and the Ideal control value needed to maintain the target trajectory in a disclosed frame format

2. The method of claim 1, wherein at least one RF metric is selected from the following: RSSI, RSCP or SIR

3. The method of claim 1, wherein at least one motion metric is selected from the following: UE horizontal velocity, UE vertical velocity and UE Bearing

4. The method of claim 1, wherein at least one packet trans- port metric is selected from the following: FER, Bit Rate, one way transmission delay

5. The method of claim 2, where at least one value is observed and recorded for a variable time interval

6. The method of claim 3, where at least one value is ob- served and recorded for a variable time interval

7. The method of claim 4, where the delay and at least one further value is observed and recorded for a time interval and the variation of delay between said observed value and the prior observed value (Jitter) is recorded

8. At least one time-variance matrix per recorded value representing observed samples in the past

9. At least one time-variance matrix per statistically estimated value representing initial predictions of such values in the future

10. A time-variance matrix combining each matrix in claim 8 in to a multi column and multi row matrix of past values. At least one value vector per category defined in claim 2, 3 and 4 is represented. Columns are measurement values and rows time intervals

11. A time-variance matrix combining each matrix in claim 9 in to a multi column and multi row matrix of future values. At least one value vector per category defined in claim 2, 3 and

4 is represented. Columns are measurment values and rows time imtervals

12. A method using the technique described in claim 11 to arrive at a prediction of a target value, with the smallest error across all possible predictions.

13. A method in which the described algorithms are used to calculate a required control input (such as power control) to achieve a given output trajectory.

14. An exchange format for the prediction results consisting of a 6 octet output frame