WO2009018585A1

WO2009018585A1 - Statistical control of radio link quality

Info

Publication number: WO2009018585A1
Application number: PCT/US2008/074425
Authority: WO
Inventors: Bijan Rohani; Kambiz Homayounfar
Original assignee: PHYBIT Pte Ltd
Current assignee: PHYBIT Pte Ltd
Priority date: 2008-08-27
Filing date: 2008-08-27
Publication date: 2009-02-05
Anticipated expiration: 2011-02-27

Abstract

Method and a system for controlling radio link quality are provided. The system includes a radio link that includes a transmitter configured to transmit bits of information through at least one medium and a receiver configured to receive the transmitted bits of information from the at least one medium. The method includes measuring a quality of a radio link, determining, in real time, an adjustment that tends to improve the measured quality of the radio link, and adjusting a parameter of the transmitter using the determined adjustment.

Description

STATISTICAL CONTROL OF RADIO LINK QUALITY

BACKGROUND OF THE INVENTION

[0001] This invention relates generally to controlling communication link quality and, more particularly, to methods and systems for real time statistical radio link quality control.

[0002] Communications over a time -varying radio channel is subject to radio channel impairments such as additive white Gaussian noise (AWGN), fading (flat and frequency-selective), and log-normal shadowing that introduce losses in the received information and degrade the quality of service (QoS) delivered. Typically, if the QoS becomes too bad, the link becomes unusable and if QoS becomes too good the link uses up resources unnecessarily. In time-varying radio channels both these scenarios can arise in the course of a connection. To ensure that the QoS for a specific application is met under varying radio channel conditions, radio link adaptation techniques are used that include a feedback loop involving radio link quality measurement and control. Measurement of the radio link quality is mainly done at the receiver and may be a measure of the received signal strength (RSS), the signal-to-noise ratio (SNR), and/or the bit-error-rate (BER) before or after the channel decoder. Control of the radio link quality includes adjustment of link parameters such as the modulation, coding, and/or power of the transmitted signal based on the radio link quality measurements.

[0003] Known adaptive coding and modulation (AMC) algorithms quantify the QoS by the minimum required BER. The BER is not an easily measured quantity. First, a bit error must be detected. This is typically accomplished by either using reference signals, or comparing the bits at the input and output of the channel decoder. Each of these methods has weaknesses which reduce their effectiveness for real time control of link quality. Second, assuming that bit errors can be detected, reliable BER measurement requires long observation times. For example, to observe 10 bit errors at a BER of 10^"3 requires observing 10⁴ bits. This is a relatively long time during which the channel could change appreciably and therefore make the measurement outdated.

[0004] In practice, the BER is not measured. Instead, the BER curves are used. These are known curves that map the channel SNR to the decoded BER. For AMC, there is a family of curves, one for each modulation and coding combination. There is a SNR value, SNR_n for each AMC scheme n, that when exceeded, the BER requirement for that scheme is met. The set of SNR_n provides switching thresholds that are used in practice for adapting AMC to channel conditions.

[0005] The BER curves used in the above scheme are pre-calculated. In practice, the mapping between the SNR and BER is influenced by many factors such as the estimation error in SNR measurement, the rate of channel change, receiver synchronization error, and equalization error. Therefore, the AMC switching thresholds cannot guarantee delivering the best AMC scheme for the existing channel condition.

BRIEF DESCRIPTION OF THE INVENTION

[0006] In one embodiment, a method for controlling radio link quality includes measuring a quality of the radio link, determining, in real time, an adjustment that tends to improve the measured quality of the radio link, and adjusting a parameter of the transmitter using the determined adjustment.

[0007] In another embodiment, a method of transmitting data through a radio link includes transmitting a plurality of data bits using a transmitter having a plurality of selectably adjustable transmitter control parameters and receiving at least some of the plurality of data bits by a receiver communicatively coupled to the transmitter through a radio link, wherein the receiver includes a link quality measurement module and an adjustment decision module. The method also includes measuring a quality of the radio link using a quality measure of the received data bits wherein the quality measure include statistical properties that indicate the stationarity of the radio link in real time, generating one or more transmitter parameter adjustment commands using the stationarity of the radio link quality measure, and adjusting a parameter of the transmitter using the generated one or more transmitter parameter adjustment commands.

[0008] In yet another embodiment, a statistical radio link quality control system includes a transmitter, a receiver communicatively coupled to the transmitter through a data transmission medium link. The system also includes a link quality measurement module communicatively coupled to the receiver. The link quality measurement module is configured to determine a quality measure of the data transmission medium link using statistical properties of the data transmission medium link wherein the statistical properties indicate a stationarity of the data transmission medium link quality. The system also includes an adjustment decision module communicatively coupled to the link quality measurement module and the transmitter. The adjustment decision module is configured to monitor the determined quality measure. The adjustment decision module is also configured to generate one or more transmitter parameter adjustment commands using the stationarity of the data transmission medium link quality, and transmit the generated one or more transmitter parameter adjustment commands to the transmitter.

BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figures 1-9 show exemplary embodiments of the methods and system described herein.

[0010] Figure IA is control chart of monitoring the state of an exemplary process xu by comparing the observed time-series against a plurality of control levels;

[0011] Figure IB is a graph of control levels for an exemplary normal probability density function (pdf);

[0012] Figure 2 is a schematic block diagram of a system for implementing statistical control of radio link quality in accordance with an exemplary embodiment of the present invention; [0013] Figure 3 is a functional block diagram illustrating a method of deriving a link quality bit error probability (BEP) from a received signal RX Signal

[0014] Figure 4 is a graph of the density function of the LLR of the 1^th bit p(λ_t) for ^5 and 10;

[0015] Figure 5 is a graph of the probability density function of p{ε_t) for γ=3;

[0016] Figure 6 is a graph of a normalized histogram of S₁ using samples of S₁ generated by using samples of λj for γ=3;

[0017] Figure 7 is a graph of the probability that S₁ falls within a particular bin shown in Figure 6;

[0018] Figure 8 is a graph 800 illustrating changes in the pdf pfø) of BEP as a result of a change in γ; and

[0019] Figure 9 is a graph illustrating changes in the pdf p(ε_t) of BEP as a result of a change in γ of plots of experimentally obtained p(ε_t).

DETAILED DESCRIPTION OF THE INVENTION

[0020] The following detailed description illustrates embodiments of the invention by way of example and not by way of limitation. It is contemplated that the invention has general application to statistical, analytical, and methodical embodiments of controlling the quality of radios links in industrial, commercial, and residential applications.

[0021] As used herein, an element or step recited in the singular and proceeded with the word "a" or "an" should be understood as not excluding plural elements or steps, unless such exclusion is explicitly recited. Furthermore, references to "one embodiment" of the present invention are not intended to be interpreted as excluding the existence of additional embodiments that also incorporate the recited features. [0022] Figure IA is control chart 100 of monitoring the state of an exemplary process xt by comparing the observed time-series against a plurality of control levels. Figure IB is a graph 102 of control levels for an exemplary normal probability density function (pdf). In the exemplary embodiment, control chart 100 is illustrated as a Shewhart chart. Control chart 100 includes an x-axis 104 graduated in units k and a y-axis 106 graduated in unit of a quality characteristic being measured Xk. A plurality of points 108 representing measurements of the quality characteristic in samples taken from the process at different times are plotted along control chart 100. A first reference line 110 is plotted at the process characteristic mean. An alarm level control limit 112 and an action level control limit 114 are plotted on control chart 100. Alarm level control limit 112 and action level control limit 114 indicate predetermined thresholds at which the process output is considered statistically becoming non- stationary and out of control respectively. Control limit 112 and 114 may be set based on standard deviations from process characteristic mean reference line 110 or may be set based on other determined or predetermined criteria. A legend 116 illustrates boundaries of areas where measurements of the quality characteristic are outside a control limit 112 and 114. Other, more powerful control schemes can be adopted depending on the needs of the process to be controlled.

LINK ADAPTATION

[0023] Communications over a time -varying radio channel is subject to radio channel impairments such as additive white Gaussian noise (AWGN), fading (flat and frequency-selective), and log-normal shadowing. Such impairments introduce losses in the received information and degrade the quality of the delivered service. Quality of service (QoS) requirements vary from one application to another. If the QoS requirement is not met exactly, QoS may degrade such that the link becomes unusable or QoS may improve to the point where the link uses up resources unnecessarily.

[0024] In time-varying radio channels both these scenarios can arise in the course of a connection. To ensure that the QoS for a specific application is met under varying radio channel conditions, radio link adaptation techniques become necessary. Such techniques include a feedback loop involving radio link quality measurement and control. Measurement of the radio link quality is mainly performed at the receiver. This entails measurement or estimation of one or more radio link measures such as the received signal strength (RSS), the signal-to-noise ratio (SNR), the bit-error-rate (BER) before or after the channel decoder. These are link variables that impact the QoS delivered on the radio link. The control part of radio link adaptation involves adapting link parameters such as the modulation, coding, and/or power of the transmitted signal within system capabilities and constraints based on the radio link quality measurements. For example, in systems such as GSM and 3G UMTS, signal modulation remain the same while channel coding and transmission power are adapted according to the prevailing channel condition to cope with quality fluctuations imposed by the time-varying mobile channel. The quality fluctuation is measured and reported by the receiver as the received signal strength and the estimated raw BER in the case of GSM, and the estimated signal-to-interference ratio (SIR) in the case of 3G UMTS.

[0025] More advanced systems based on standards such as WiMAX, 3G LTE, and IEEE 802.11a have incorporated variable multilevel modulation schemes such as the quadrate amplitude modulation (QAM) in addition to variable coding and transmission power. This allows the added flexibility of adapting the modulation order to the channel condition such that progressively higher data rates can be used under better channel conditions. In this way, a better spectral efficiency in bits/s/Hz can be achieved in time-varying mobile channels. In addition the channel coding can be adjusted to maintain a given level of QoS. Such adaptive coding and modulation (AMC) schemes require sophisticated algorithms to decide the best combination of modulation and coding schemes from an available set of schemes - AMC set - that not only optimizes the channel spectral efficiency but also ensures delivery of the required QoS based on the prevailing channel conditions.

[0026] The current state-of-the-art AMC algorithms quantify the QoS by the minimum required BER. The BER is not an easily measured quantity because of the following reasons. First, a bit error must be detected. This can be done by either using reference signals, or comparing the bits at the input and output of the channel decoder, but such methods have weaknesses which makes them difficult and/or unreliable to use. Second, assuming that bit errors can be detected, reliable BER measurement requires long observation times. For example, to observe 10 bit errors at a BER of 10^~3 requires observing 10⁴ bits. This is a relatively long time during which the channel could change appreciably and therefore make the measurement outdated.

[0027] In a more practical approach, the BER is not measured. Instead, the BER curves of the allowed AMC set are used. These are known curves that map the channel SNR to the decoded BER. For AMC, there is a family of curves, one for each modulation and coding combination. There is a SNR value, SNR_n for each AMC scheme n, that when exceeded the BER requirement for that scheme is met. The set of SNR_n provides switching thresholds that are used in practice for adapting AMC to channel conditions.

[0028] The BER curves used in the above scheme are pre-calculated. In practice, the mapping between the SNR and BER is determined by many factors such as the estimation error in SNR measurement, the rate of channel change, receiver synchronization error, and equalization error. Therefore, the AMC switching thresholds cannot guarantee delivering the best AMC scheme for the existing channel condition.

STATISTICAL PROCESS CONTROL

[0029] Statistical process control (SPC) is used in industries such as manufacturing and chemical engineering for monitoring and control of sophisticated processes. In SPC, the output of a process is viewed as being random in nature varying in accordance with some underlying statistical model. SPC provides powerful tools for monitoring and control of processes based on their underlying statistics. In the SPC context a process is generally considered to be in either of two states: under control or out of control. When in control, the process is only affected by common causes. Common causes cannot be removed, and the process variations in this state are only due to these common causes. The process in this state is stationary. When out of control, the process is affected by special causes. The process variations are due to both common and special causes in this state. The process is non-stationary in this state. In order to restore the process to the state of control, special causes must be identified and removed.

[0030] An example of a typical random process is shown in Figure IA. Samples 108 of the process have been represented by the time-series Xk plotted against the time index k. In the exemplary embodiment, xt is assumed to be normally distributed as shown in Figure IB with its pdf given by p(x_k). In an alternative embodiment, the process pdf is not normal. With reference to Figure IB, xt vary randomly around the process mean μ_x when the process in a state of control. In this state only common causes are present and the process is stationary. In this state, the sample values xt should lie in the interval μ_x±2σ_x with a probability of 0.954, and fall in the interval μ_x±3σ_x with an even higher probability of 0.997. That is to say, if the process is in a state of control, its sample values must almost certainly fall within the μ_x±3σ_x range. The ±3σ_x action level control limits 114 on the process variations define the boundaries beyond which the process is determined to be out of control and non-stationary. Action must be taken in order to identify the special causes and remove them so that the state of control can be restored. The ±2σ_x alarm level control limits 112 are also useful for process monitoring. Alarm level control limits 112 are warning levels and are used as alarms that the process is showing signs of becoming non-stationary.

[0031] Figure 2 is a schematic block diagram of a system 200 for implementing statistical control of radio link quality in accordance with an exemplary embodiment of the present invention. Statistical radio link quality control (SRLQC) is a framework for the design of AMC control algorithms. A difference between SRLQC and other known methods is that SRLQC does not rely on SNR measurement or on the accuracy of mapping between SNR and BER to determine the best AMC scheme for the channel condition. STATISTICAL CONTROL OF RADIO LINK QUALITY

[0032] With the brief background given on the SPC, it is possible to examine the function of radio link adaptation in the context of a statistical process control problem. The radio link quality is random in nature and can be represented by a random time-series. When the link quality is under control, its variations are due to common causes, such as AWGN and hardware impairments, that cannot be removed. The time-varying channel gain due to multipath fading and shadowing, which constitute special cases, will have been removed in this state. The underlying statistical model in this case is that of the link quality in an AWGN channel. The link quality is stationary and samples vary around a constant mean value with a constant variance. The variations of the quality can be modeled by a pdf reflecting the receiver performance in an AWGN channel. The pdf parameters μ and σ are constants that characterize the link quality when it is under control.

[0033] Conversely, when the link quality is out of control, channel gain variations due to fading and shadowing are present. As such, the link quality varies under the influence of these factors as well as AWGN. The link quality becomes non-stationary in this state. The non-stationarity can be better understood if the channel can be approximated as being quasi-static. In such a scenario, the transmission time is divided into intervals in which the channel gain is considered to be constant within each interval, and changing from interval to interval. As such, the channel in each time interval behaves like an AWGN channel whose SNR depends on the channel gain in that interval. The pdf that models the link quality variations changes from one time interval to the next because of changing SNR. This gives rise to variations in the mean-value and variance of the link quality with time, thus non- stationarity in quality.

[0034] A block diagram showing the structure for implementing statistical control of radio link quality is shown in Figure 2. TX Bits 202 represent the information bits to be transmitted. The information could be audio, video, text, data, or a combination thereof. The functional block Transmitter 204 represents a chain of physical layer functions that are performed on TX Bits 202 to prepare them for transmission over a radio link 206. Such physical layer functions include, for example, but not limited to source coding, interleaving/channel coding, modulation, filtering and amplification. The specific parameters for the physical layer functions vary according to the wireless standard adopted for implementing system 200. The same principles for SRLQC apply regardless of the adopted standards. Link 206 represents a radio channel. Link 206 provides a medium for the flow of information from Transmitter 204 to a Receiver 208 of a receiving system 209. A radio channel can introduce a variety of impairments on the transmitted signal that may lead to loss of parts or all of the transmitted information. The channel embodies AWGN, multipath fading, and shadowing. It is also customary to lump receiver impairments such as synchronization errors, and I&Q imbalance into the channel.

[0035] The functional block Receiver 208 is responsible for recovering TX bits 202 from the received signal on Link 206. It does so by applying the inverse physical layer functions corresponding to those applied in Transmitter 204 to generate corresponding RX bits 210. In an ideal case, the recovered bits RX bits 210 are identical to TX Bits 202. In practice, however, a fraction of RX Bits 210 is received in error leading to some information loss. The degree of loss that can be tolerated depends on the type of the transmitted information. For example, a BER of 10^"2 can be acceptable for voice while any error in most data transmissions is regarded unacceptable. The degree of loss introduced by Link 206 is quantified by a Quality Measure module 212. In particular, in a time varying radio channel, the measured quality fluctuates with time. In such cases, link quality is stabilized by determining adequate adjustments to be carried out at transmitter 204. The adjustment determination is implemented by an Adjustment Decision module 214. The decisions are sent to Transmitter 204 as an Adjustment Command 216 on a feedback path 218.

[0036] Figure 3 is a functional block diagram illustrating a method 300 of deriving a link quality BEP 302 from a received signal RX Signal 304. Quality Measure 212 plays an important part in SRLQC. Quality Measure 212 is closely related to the QoS delivered by RX Bits 210. In addition, Quality Measure 212 is measured in real-time. An example of Quality Measure 212 is the bit error probability (BEP). BEP can be estimated from the log likelihood ratios (LLR) of a soft-decision decoder 306, which is a common feature of the modern wireless standards.

[0037] A Demodulator 308 of receiver 208 (shown in Figure 2) maps RX Signal 304 to RX Symbols 310. RX Symbols 310 are then decoded by a soft- decision algorithm 312 represented by function block Decoder 306 to derive the output bits RX Bits 210. Subsequently, the block Quality Measure 212 determines BEP 302 from log-likelihood ratios LLR 314, which is a readily available byproduct of soft-decision algorithm 312 in Decoder 306.

STATISTICAL MODEL OF BEP

[0038] Let X₁ denote the LLR for the zth bit in sequence of decoded RX Bits 210. By definition, X₁ represents the likelihood that the decoded bit is correct. Therefore, the probability of error S₁ for the decoded bit, i.e. BEP, is calculated according to:

[0039] Equation (1) provides a way to determine the radio link quality based on information that is already available in real-time at receiver 208 (shown in Figure 2), for example, LLR 314 from decoder 306.

[0040] A density function p{Sj) represents the underlying statistical model governing variations of S₁ for binary phase-shift keying (BPSK) signaling in an AWGN channel with bit-energy-to-noise-spectral density of χ = E_b/N_o . It is assumed that the BPSK signal is represented by constellation points U₁ = +-^E_b ^~ . In this case, the received signal at the output of a matched filter is given by D₁ = U₁ + n_t where ni is a sample of AWGN with spectral density of N₀Il W/Hz. The corresponding LLR is given by:

λ =- -(U₁ + n_;) (2)

N [0041] Accordingly, from equation (2) the pdf P(AJu₁) of LLRs conditioned on the transmitted signal w is determined to be:

= N(4_rsgn(_Ul),8r)

[0042] Here, ΛT(_//,σ²) represents normal distribution with a mean- value of// and a variance of σ². Additionally, sgn(x) is the signum function.

[0043] Assuming p(u, =

= 0.5, then

P(A₁) = 0.5N(-4r,8χ)+0.5N(4r,8χ) (4)

[0044] Figure 4 is a graph 400 of the density function of the LLR of the ith bit p(λ_t) for γ=5 and 10. A first trace 402 corresponds to γ=5 and a second trace 404 corresponds to a γ=\Q. The pdf for each value of γ comprises a pair of normal distributions centered at —Aγ and +Aγ. The variance of each distribution is Sχ.

[0045] With p(λ_t) given by equation (4), the pdf p(ε_t) of BEP is determined by applying a change of variable according to equation (1). The resulting pdf is given by:

[0046] It should be noted that p(ε_t) is non-zero only within the interval 0 < ε_t < 0.5.

[0047] Figure 5 is a graph 500 of the probability density function of p(ε,) for γ=3. Graph 500 includes a trace 502 of probability density function p(ε,) that resembles a delta function achieving a very large peak 504 at an extremely small BEP £i. [0048] Figure 6 is a graph 600 of a normalized histogram of S₁ using samples of S₁ generated by using samples of λj for γ=3. Graph 600 includes an x-axis 602 graduated in units of S₁ up to approximately 2x10 ⁹ and a y-axis 604 graduated in units of p{ε) Equation 5 was verified numerically by first generating samples of λi according to equation 4 for γ=i. The generated samples of λi were then used to generate samples of S₁ according to equation 1. The histogram of S₁ was obtained and normalized to the total number of samples. Note that, the pdf has been shown only for S₁ <2xlO^"9 in order to observe the pdf around its peak value clearly.

[0049] By definition, the histogram height at the centre of each bin 606 represents the estimate of the probability that S₁ falls within that particular bin. This probability is calculated by integrating the expression forpfø) in equation 5 over the interval defined by the histogram bin.

[0050] Figure 7 is a graph 700 of the probability that S₁ falls within a particular bin 606 (shown in Figure 6). In the exemplary embodiment, the probability is calculated by integrating the expression forpfø) in equation 5 over the interval defined by the histogram bin for the histogram of Figure 6.

[0051] Comparison of Figure 6 and 7 illustrates that the theoretical results calculated from equation 5 are consistent with numerically generated results given by the normalized histogram. The same method can be used to verify equation 5 for values of pother than 3. The effect of increasing EjJN₀ on pfø) is demonstrated as follows. First, the value ε^* of BEP for which the peak occurs is approximate by evaluating the expression for BEP in equation 1 at the mean- value of

. This gives:

ε ∞ e-*' (6)

[0052] The error of approximation decreases as γ increases. The peak value of p(ε_t) is approximated to be:

p(ε) « -f= (7) [0053] As evidenced by equations 6 and 7, p{ε_t) peaks at smaller values of BEP while achieving larger peaks as EbIN₀ increases.

[0054] Figure 8 is a graph 800 illustrating changes in the pdf pfø) of BEP as a result of a change in γ. The probability density functions in Figure 8 are calculated for γ₂>γi based on the expression of pfø) in equation 5. We have used logarithmic scale for ει so the peaks of the two probability density functions can be seen clearly. It is observed from Figure 8 that the peak value of the pdf occurred at a smaller value of BEP while a larger peak was achieved as EtJN₀ was increased.

[0055] The dependence of p(ε_t) on EJN₀ can be confirmed by simulation. For example, the LLR values of a BPSK signal at the output of demodulator 308 and decoder 306 for Y₁=I l dB and γ₂ ⁼ 12 dB are obtained. The corresponding BEP values are calculated according to equation 1 and the results are used for estimation of p(ε_t) for Y₁=I 1 dB and γ₂ ⁼ 12 dB.

[0056] Figure 9 is a graph 900 illustrating changes in the pdf pfø ) of BEP as a result of a change in γ of plots of experimentally obtained p(ε_t). In Figure 9, a change in the location and value of the pdf peak is clearly observable . Although a derivation of p{ε_t) for BPSK signaling was described above, it should be noted that a similar derivation for other other modulation schemes such as, quadrature phase-shift keying (QPSK), 8 Phase Shift Keying (8PSK), and M-ary quadrature amplitude modulation (M-QAM) can also be shown. In addition the derivation may be for P(JzIj) instead of p(ε_t).

IMPLEMENTATION OF ADJUSTMENT DECISION WITH SPC

[0057] With reference to Figure 2, the Quality Measure 212 S₁ is used as the input to Adjustment Decision 214 of the SRLQC process. Adjustment Decision 214 is an SPC unit that implements the functions of monitoring the stationarity of Quality Measure 212, and making appropriate decisions based on the outcome of the monitoring according to the following three rules. If Link 206 is stationary, no adjustment is made. If Link 206 is approaching non-stationarity, a minor adjustment is made, and if Link 206 is already non-stationary, a major adjustment is made or Link 206 is disconnected. These three rules comprise the link adaptation policies of SRLQC. The first rule states that the radio link quality is satisfactory and no changes to the current link parameters are necessary. An adjustment command is not transmitted to the transmitter or a non-adjustment command is sent to the transmitter wherein the non-adjustment command indicates the feedback link is still operable but that no changes to the current link parameters are necessary. The second rule is applied when the quality is still adequate but there are signs that it is becoming non- stationary. This could mean that the quality is becoming much better than required or it is approaching its limit of becoming unacceptably poor. This scenario is very likely in a time-varying radio channel and it is possible that the second rule has to be carried out quite often. In this case, a minor adjustment can be applied to the link. This may be done, for example, by adjusting the transmitter power up or down in small increments. This allows link adaptation in a fast and efficient way. As used herein, the power down adjustment is applied when the link is becoming better than required and is associated with a Part I of the second rule, and the power up adjustment is applied when the link is approaching its limit of becoming unacceptably poor and is associated with a Part II of the second rule.

[0058] The link adaptation policy presented by the second rule allows only for small link quality adjustments within a given AMC scheme. More severe scenarios, whereby link quality cannot be handled by merely applying small adjustments, require major changes to the link. The third rule embodies the link adaptation policy for such scenarios. In this case, a major adjustment is applied to the link parameters. For example, a different AMC scheme is adopted for the signal transmission. The selection of the AMC scheme should allow the highest possible modulation order while satisfying the link quality requirements. Progressively more robust AMC schemes are selected as the link quality deteriorates. Conversely, as the link quality improves AMC schemes with increasingly higher throughput are selected. Although it is possible in the implementation of the third rule to allow jumps between any two permissible schemes within the AMC set, the channel does not change so abruptly to necessitate that. In practice, the next higher or lower AMC scheme to the current one is chosen for more throughput or more robust transmission, respectively. In the event that the most robust AMC is already in use and the quality requirement is not met the link can be disconnected. As used herein, the higher AMC selection is applied to improve throughput and is associated with a Part I of the third rule, and the lower AMC selection is applied as the link quality deteriorates and is associated with a Part II of the third rule.

[0059] Adjustment Decision 214 embodies a SPC algorithm for monitoring the stationarity of Quality Measure 212. The timely application of the adjustment rules and consequently the performance of the SRLQC scheme depend on this SPC algorithm. There are several algorithms available in the literature that can be adopted for SRLQC. These include the Shewhart chart, exponentially weighted moving average (EWMA), and the cumulative sum (cusum) scheme. Although, the implementation based on the cusum scheme is further described below, it should be noted that any one of the SPC monitoring schemes can be incorporated for SRLQC. The cusum scheme is particularly powerful and sensitive for detecting small deviations in the pdf of the process that is being controlled.

[0060] A so-called two-sided cusum scheme can be used for monitoring of the link quality. In this case, upper and lower cusums QiAf) ^and Qdf) are determined per observed S₁ according to the following recursive expressions,

Q_L(i)=min(0, S₁ -T+Q_L(i- 1 )) (7b)

[0061] where T is the so-called target value for S₁.

[0062] Q_H is used for monitoring an increase in the BEP, i.e. a degradation in the link quality, and Q_L is used for monitoring the reverse situation i.e. an improvement in the link quality.

[0063] The initial values of Q_H and Q_L can be chosen to be zero. However, optimal values for the initialization of the algorithm can be calculated which lead to faster detection speeds. [0064] The cusum equations of (7a) and (7b) essentially measure the accumulated deviation of S₁ from its target value T over a period. When the link quality is under control, the deviations of S₁ on both sides of the target on average cancel out, hence Q_H and Q_L stay near zero. It is possible to observe a consecutive run of S₁ values, which fall on the same side of the target value. In such events, S₁ -T has the same sign for a consecutive run of observations leading to accumulation of Q_H or Q_L in one direction - Q_H keeps increasing or Q_L keeps decreasing. Generally, large deviations of Q_H or Q_L from zero are rare, as that would require longer runs of S₁ -T with the same sign. The run lengths of S₁ occur with probabilities that can be exactly determined from the pdf P(S₁). In other words, for a stationary process, values of Q_H and Q_L occur with definite probabilities. In the same way that warning and action levels can be calculated for S₁ given p(Sj), corresponding levels can also be calculated for Q_H and Q_L based on P(S₁).

[0065] Let V_L and rπ denote the warning limits for the lower and upper cusums, respectively. Also, let H_L and hπ denote the corresponding action limits. Then, the SRLQC algorithm for Adjustment Decision is summarized below.

if QH(Ϊ)< r_H and Q_L(i)> r_L apply the first rule: don't do anything elseif Oz(0< rz

apply part I of the second rule: request for decreased transmission power else apply part I of the third rule: request for next higher AMC scheme dsQif QH(Ϊ)> r_H iϊQdϊ)< h_H apply part II of the second rule: request for increased transmission power else apply part II of the third rule: request for next lower AMC scheme

[0066] The performance of the cusum scheme is measured in terms of range of tolerable link quality variation, the delay in detection of a change in p(si), and the probability of false alarm. The detection delay is the average number of samples (run length) of S₁ that are observed from the moment a change in p{ε_t) occurs until the change is detected by Q_H or Q_L crossing a warning or action limit. The probability of false alarm refers to the non-zero probability that Q_H or Q_L can cross a warning or action limit even though P(S₁) remains unchanged. There is an underlying interrelationship among these performance measures, which necessitates careful design of algorithm parameters T, r_L, r_H, H_L and hπ- For example, the detection delay cannot be decreased without increasing the probability of false alarm or decreasing the allowable range of quality variation.

[0067] The cusum algorithm design for the Gaussian pdf has been extensively documented in the published literature. However, the P(S₁) given by equation 5 is non-Gaussian. In this case, the cusum algorithm can be designed in at least two possible ways regardless of the expression for the pdf. The first method, calculates the performance measures using a numerical approach. In the second method, consecutive samples of S₁ can be combined - either through summation or averaging - to generate a new time-series. If a sufficiently large number of samples are combined in this way, the new time-series will have a Gaussian pdf owing to the Central Limit Theorem. Thus, the standard cusum design for Gaussian pdf can be applied to resulting time-series.

[0068] The term processor, as used herein, refers to central processing units, microprocessors, microcontrollers, reduced instruction set circuits (RISC), application specific integrated circuits (ASIC), logic circuits, and any other circuit or processor capable of executing the functions described herein.

[0069] As used herein, the terms "software" and "firmware" are interchangeable, and include any computer program stored in memory for execution by a processor, including RAM memory, ROM memory, EPROM memory, EEPROM memory, and non-volatile RAM (NVRAM) memory. The above memory types are exemplary only, and are thus not limiting as to the types of memory usable for storage of a computer program. [0070] As will be appreciated based on the foregoing specification, the above-described embodiments of the disclosure may be implemented using computer programming or engineering techniques including computer software, firmware, hardware or any combination or subset thereof, wherein the technical effect is automatically adjusting transmitter parameters in real time based on a measure of a quality of the communication link. Any such resulting program, having computer- readable code means, may be embodied or provided within one or more computer- readable media, thereby making a computer program product, i.e., an article of manufacture, according to the discussed embodiments of the disclosure. The computer readable media may be, for example, but is not limited to, a fixed (hard) drive, diskette, optical disk, magnetic tape, semiconductor memory such as read-only memory (ROM), and/or any transmitting/receiving medium such as the Internet or other communication network or link. The article of manufacture containing the computer code may be made and/or used by executing the code directly from one medium, by copying the code from one medium to another medium, or by transmitting the code over a network.

[0071] The above-described embodiments of a methods and system of controlling a quality of a communication link in real time provides a cost-effective and reliable means for measuring the quality using soft-decision decoder LLR values and adjusting parameters of the transmitter based on the measured quality. As a result, the methods and system described herein facilitate improving a quality of service of a communication link in a cost-effective and reliable manner.

[0072] While the disclosure has been described in terms of various specific embodiments, it will be recognized that the disclosure can be practiced with modification within the spirit and scope of the claims.

Claims

WHAT IS CLAIMED IS:

1. A method of controlling radio link quality wherein a radio link includes a transmitter configured to transmit bits of information through at least one medium and a receiver configured to receive the transmitted bits of information from the at least one medium, said method comprising:

measuring a quality of a radio link;

determining, in real time, an adjustment that tends to improve the measured quality of the radio link; and

adjusting a parameter of the transmitter using the determined adjustment.

2. A method in accordance with Claim 1 wherein measuring a quality of a radio link comprises measuring the quality of the radio link by the receiver.

3. A method in accordance with Claim 1 wherein measuring a quality of a radio link comprises measuring a bit error probability (BEP) of the radio link.

4. A method in accordance with Claim 3 wherein measuring a bit error probability (BEP) of the radio link comprises estimating the bit error probability (BEP) of the radio link using a log likelihood ratios (LLR) of a soft-decision decoder in the receiver.

5. A method in accordance with Claim 1 wherein adjusting a parameter of the transmitter using the determined adjustment comprises transmitting an adjustment command from the receiver to the transmitter.

6. A method in accordance with Claim 1 further comprising transmitting an adjustment command from the receiver to the transmitter using a feedback link.

7. A method in accordance with Claim 6 wherein transmitting an adjustment command from the receiver to the transmitter using a feedback link comprises transmitting an adjustment command from the receiver to the transmitter using a feedback link that is separate from the radio link.

8. A method in accordance with Claim 1 wherein measuring a quality of a radio link comprises determining a quality measure of the radio link wherein the radio link comprises a time-series having statistical properties indicate the stationarity of the radio link quality.

9. A method in accordance with Claim 8 wherein determining a quality measure of the radio link that includes a time-series having statistical properties indicate the stationarity of the radio link quality comprises determining whether the link is at least one of stationary, becoming non-stationary, and is already non-stationary.

10. A method in accordance with Claim 8 wherein determining a quality measure of the radio link comprises determining at least one of a received signal strength (RSS), a signal-to-noise ratio (SNR), a bit-error-rate (BER) before a channel decoder, and a bit-error-rate (BER) after the channel decoder.

11. A method in accordance with Claim 8 wherein determining an adjustment comprises:

monitoring the quality measure for stationarity of the radio link quality;

if the radio link quality is non-stationary, transmitting a major adjustment command to the transmitter;

if the radio link quality is approaching non-stationarity, transmitting a minor adjustment command to the transmitter; and if the radio link quality is stationary, at least one of not transmitting an adjustment command to the transmitter and transmitting a non-adjustment command to the transmitter.

12. A method in accordance with Claim 1 wherein adjusting a parameter of the transmitter comprises adjusting at least one of a modulation, coding, and power of the transmitter based on the radio link quality measurements.

13. A method of transmitting data through a communication link, said method comprising:

transmitting a plurality of data bits using a transmitter having a plurality of selectably adjustable transmitter control parameters;

receiving at least some of the plurality of data bits by a receiver communicatively coupled to the transmitter through a communication link, wherein the receiver includes a link quality measurement module and an adjustment decision module;

measuring a quality of the communication link using a quality measure of the received data bits wherein the quality measure include statistical properties that indicate the stationarity of the communication link in real time;

generating one or more transmitter parameter adjustment commands using the stationarity of the communication link quality measure; and

adjusting a parameter of the transmitter using the generated one or more transmitter parameter adjustment commands.

14. A method in accordance with Claim 13 wherein measuring a quality of the communication link comprises estimating a bit error probability (BEP) of the communication link using a log likelihood ratios (LLR) of a soft-decision decoder in the receiver.

15. A statistical communication link quality control system comprising: a transmitter; a receiver communicatively coupled to the transmitter through a data transmission medium link; a link quality measurement module communicatively coupled to said receiver, said link quality measurement module configured to determine a quality measure of said data transmission medium link using statistical properties of said data transmission medium link wherein the statistical properties indicate a stationarity of the data transmission medium link quality; and an adjustment decision module communicatively coupled to said link quality measurement module and said transmitter, said adjustment decision module configured to: monitor the determined quality measure; generate one or more transmitter parameter adjustment commands using the stationarity of the data transmission medium link quality; transmit the generated one or more transmitter parameter adjustment commands to said transmitter.

16. A system in accordance with Claim 15 wherein said receiver comprises:

a demodulator configured to map a received RX signal to RX symbols; a channel decoder configured to decode the RX symbols using a soft- decision algorithm used to generate output RX bits, said decoder further configured to determine a bit error probability from a log-likelihood ratio generated by the soft- decision algorithm.

17. A system in accordance with Claim 15 wherein said adjustment decision module is further configured to monitor the quality measure for stationarity of the communication link quality.

18. A system in accordance with Claim 15 wherein said adjustment decision module is further configured to: transmit a major adjustment command to the transmitter if the communication link quality is non-stationary;

transmit a minor adjustment command to the transmitter if the communication link quality is approaching non-stationarity; and

at least one of transmit no adjustment command to the transmitter and transmit a non-adjustment command to the transmitter if the communication link quality is stationary.

19. A system in accordance with Claim 15 wherein said quality measure comprises at least one of a received signal strength (RSS), a signal-to-noise ratio (SNR), a bit-error-rate (BER) before a channel decoder, and a bit-error-rate (BER) after the channel decoder.

20. A system in accordance with Claim 15 wherein said quality measure is closely related to the quality of service delivered by the output RX Bits.