Method for adapting a link for selecting a frame transmission mode and corresponding WI-FI access point

Abstract

A link adaptation method uses a selection criterion for selecting a transmission mode for transmitting PPDU frames over a channel of a telecommunications system. The system includes an access point and a plurality of stations having various different transmission modes associated with different data rates. Access to the channel being of the random type. The selection makes use of a time occupancy metric γ of the channel.

Claims

1. A link adaptation method comprising: using a selection criterion for selecting a transmission mode for transmitting physical layer protocol data unit (PPDU) frames to be transmitted over a channel of a telecommunications system having an access point and at least one station possessing various transmission modes associated with different instantaneous bit rates, access to the channel being of the random type, wherein selecting a transmission mode makes use of a time occupancy metric γ for the channel that is expressed in the following form: $\gamma =10.Math.{\log }_{10}\left(\frac{t_{\mathrm{acc}}}{t_{{\mathrm{tra}}_m}}\right)$ with t.sub.acc being the mean access time to the channel over the observation window, t.sub.tra.sub.m being the time needed to transit a PPDU frame using the transmission mode m, and being expressed in the following form: $t_{{\mathrm{tra}}_m}=t_{\mathrm{header}}+\frac{\mathrm{Data}}{\mathrm{Datarate}\left(m\right)}W$ with Data being the load in bytes of a physical layer service data unit (PSDU) for transmitting in the PPDU frame, and Datarate being the instantaneous data rate corresponding to the transmission mode m being used.

2. The link adaptation method according to claim 1, wherein the time determination for determining the mean access time to the channel t.sub.acc is performed from a transmission mean data rate.

3. The link adaptation method according to claim 2, wherein the mean data rate is expressed as a function of the probability of successfully transmitting the PPDU frames associated with a service n, said probability of success P.sub.S(G) being estimated as follows: $P_s\left(G\right)•\frac{1}{1+G}$ with G being the occupancy rate of the link.

4. The link adaptation method according to claim 1 comprising: comparing a threshold value with the values of the metric γ as determined for different transmission modes in order to preselect modes; and selecting in compliance with a determined criterion of one mode from among the various preselected modes.

5. The link adaptation method according to claim 1, wherein occupancy of the channel is measured over a determined observation window and makes use of a Poisson distribution model for the arrival rates of the PPDU frames to be transmitted, the variance λ of the Poisson distribution corresponding to the mean frequency of arrival of the PPDU frames.

6. The link adaptation method according to claim 1, wherein the various transmission modes are classified by groups of equivalent data rates D, and for a group, the method comprises: determining the value of a first metric α for a given transmission mode, which metric measures relative degradation of the communication signal at a given distance d as introduced by the transmission medium for a given environment relative to a reference model of the transmission medium, resulting from a multi-path effect and/or an attenuation effect of the transmission medium; and comparing values of the first metric α for different modes for selection with at least one transmission mode (Mode.sub.i) per group.

7. A WiFi access point comprising: M transmission modes that are associated with respective different instantaneous data rates, M≧2, for a telecommunications system having a plurality of stations and the access point, access to the transmission channel of the WiFi telecommunications system being of the random type; a non-transitory computer-readable medium comprising instructions stored thereon; a processor configured by the instructions to perform acts comprising: determining the value of a first time occupancy metric γ.sub.n,m of the channel for a given transmission mode, which metric is expressed in the following form: $\gamma =10.Math.{\log }_{10}\left(\frac{t_{\mathrm{acc}}}{t_{{\mathrm{tra}}_m}}\right)$  with t.sub.acc being the mean access time to the channel over the observation window, t.sub.tra.sub.m being the time needed for transmitting a physical layer protocol data unit (PPDU) frame using transmission mode m, which time is expressed in the following form: $t_{{\mathrm{tra}}_m}=t_{\mathrm{header}}+\frac{\mathrm{Data}}{\mathrm{Datarate}\left(m\right)}$  with Data being the load in bytes of a physical layer service data unit (PSDU) to be transmitted in the PPDU frame, Datarate being the instantaneous data rate corresponding to the transmission mode m in use; comparing the values of the metric γ.sub.n,m for different modes with a threshold value in order to preselect transmission modes; and making a selection in compliance with a determined criterion from among the various preselected modes.

8. A telecommunications system including a WiFi access point according to claim 7.

9. A non-transitory computer-readable data medium including program instructions adapted to perform a link adaptation method, when said program is loaded and executed in a WiFi access point the link adaptation method comprising: using a selection criterion for selecting a transmission mode for transmitting physical layer protocol data unit (PPDU) frames to be transmitted over a channel of a telecommunications system having an access point and at least one station possessing various transmission modes associated with different instantaneous bit rates, access to the channel being of the random type, wherein selecting a transmission mode makes use of a time occupancy metric γ for the channel that is expressed in the following form: $\gamma =10.Math.{\log }_{10}\left(\frac{t_{\mathrm{acc}}}{t_{{\mathrm{tra}}_m}}\right)$ with t.sub.acc being the mean access time to the channel over the observation window, t.sub.tra.sub.m being the time needed to transit a PPDU frame using the transmission mode m, and being expressed in the following form: $t_{{\mathrm{tra}}_m}=t_{\mathrm{header}}+\frac{\mathrm{Data}}{\mathrm{Datarate}\left(m\right)}\underset{\_}{W}$ with Data being the load in bytes of a physical layer service data unit (PSDU) for transmitting in the PPDU frame, and Datarate being the instantaneous data rate corresponding to the transmission mode m being used.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

(1) Other characteristics and advantages of the invention appear from the following description of particular examples made with reference to the accompanying figures given as non-limiting examples.

(2) FIG. 1 is a diagrammatic representation of the structure of a frame of the physical layer at level 1 and of the corresponding frame of the MAC layer at level 2, with reference to the OSI model.

(3) FIG. 2 is a diagram showing the virtual exchange of frames between the physical layers PHY and MAC of an emitter EM and of a receiver RE, and it shows an example of SDU data being encapsulated in the PDUs for the PHY and MAC layers ({PSDU and PPDU}) for the PHY layer, {MSDU,MPDU} for the MAC layer).

(4) FIG. 3 is a diagram showing the times during which the transmission channel is occupied by frames.

(5) FIG. 4 is a diagram showing how the mean access time t.sub.acc to the channel is determined by observing frames transmitted over the channel so as to be able to estimate the parameter λ, X.sub.n, and X, λ being the arrival rate of estimated frames at the access point over the time window t.sub.obs and X being the mean duration of frames over the same observation duration.

(6) FIG. 5 is a flow chart of the main steps for performing a first implementation of the method of the invention.

(7) FIG. 6 is a flow chart showing how the step of selecting an interface and a transmission mode of FIG. 5 takes place in a first implementation of the method that uses a GLB link adaptation algorithm, i.e. in a multi-technology interface context.

(8) FIG. 7 is a flow chart showing how the correction step of FIG. 5 takes place in a second implementation of the method that uses a MiSer link adaptation algorithm.

(9) FIG. 8 is a diagram of a single technology implementation of the invention with a WiFi type communication system having an access point AP and two stations STA.sub.1 and STA.sub.2.

(10) FIG. 9 shows the attenuation of the propagation channel at 5 gigahertz (GHz).

(11) FIG. 10 is a simplified diagram showing the structure of an access point of the invention.

DETAILED DESCRIPTION OF ILLUSTRATIVE EMBODIMENTS

(12) The invention is described while considering a particular use of a so-called WiFi wireless communication system that has one access point and two stations in communication with the access point. The transmission channel, which has access of random type, is shared between communication entities, i.e. the access point and the stations. It is also considered that only one service is activated by a user and that a station is used by only one user. This is naturally an example that is illustrative and not limiting, and it may be generalized to more than two stations, to more than one service per station, and to more than one user per station, and such generalization of the example that is described lies within the competence of the person skilled in the art, given the following description.

(13) In the example, the access point and the stations are compatible with IEEE standard 802.11_2012. The access point and the stations have various transmission modes, which are associated with various different instantaneous data rates.

(14) The method of the invention is performed by the access point. The purpose of the proposed method is to determine the most efficient way of transmitting data relating to a destination service from the first station while complying with a mean data rate D.sub.moy and a QoS for this station, and while taking account of the occupancy of the channel by the data transmitted between the access point and the second station.

(15) Given the random access to the transmission channel, traffic on the channel as shown in FIG. 3 may be modelled using a Poisson distribution of parameter λ.Math.t.sub.OBS where λ is the arrival frequency of N frames over an observation window t.sub.OBS for the WiFi system as a whole:

(16) $\begin{array}{cc}P\left\{X\left(t+t_{OBS}\right)-X\left(t\right)=N\right\}=e^{-\lambda .Math.t_{OBS}}.Math.\frac{{\left(\lambda .Math.t_{OBS}\right)}^N}{N!}& \left(2\right)\end{array}$

(17) The invention uses a metric γ for evaluating the time occupancy of the transmission channel in order to select the transmission mode and possibly the interface that presents the greatest energy efficiency for transmitting a frame that satisfies the given constraints. The energy efficiency of a transmission mode associated with an interface and a transmission mode is the ratio of the mean data rate to the energy needed for emitting in order to transmit the service, i.e. the required emitting power integrated over a time window T (corresponding to the observation duration for determining the mean data rate) in order to deliver the data rate while complying with the QoS constraint (target binary or packet error rate (BER, PER).

(18) This metric γ corresponds to a time spreading coefficient that represents the ratio of the transmission channel availability time over an observation period t.sub.OBS to the time needed for transmitting a frame t.sub.tra. The transmission channel availability time in a random access context is evaluated by the mean access time t.sub.acc to the communication channel. The metric γ is expressed as follows:

(19) $\begin{array}{cc}\gamma =10.Math.{\log }_{10}\left(\frac{t_{\mathrm{acc}}}{t_{\mathrm{tra}}}\right)& \left(3\right)\end{array}$

(20) The occupancy of the channel is expressed in the following form:
G=λ.Math.X  (4)
with:

(21) $\begin{array}{cc}E\left\{t_n\right\}=\frac{1}{\lambda }& \left(5\right)\\ E\left\{X_n\right\}=\stackrel{\_}{X}& \left(6\right)\end{array}$
where the operator E{x} represents the expectation of the variable x and X is the mean time duration of the PPDU frames transmitted over the channel.

(22) The mean access point t.sub.acc of the transmission channel is then given by the following formula:

(23) $\begin{array}{cc}{t}_{\mathrm{acc}}=E\left\{t_n-X_n\right\}=\frac{1}{\lambda }-\stackrel{\_}{X}& \left(7\right)\end{array}$

(24) By combining equations (4) and (7), the following is obtained:

(25) $\begin{array}{cc}{t}_{\mathrm{acc}}=\frac{1-\lambda \stackrel{\_}{X}}{\lambda }=\frac{1-G}{\lambda }& \left(8\right)\end{array}$

(26) The time t.sub.tra needed for transmitting a frame depends on the transmission mode m that is used, and is expressed as follows:

(27) 0$\begin{array}{cc}{t}_{{\mathrm{tra}}_m}=t_{\mathrm{header}}+\frac{\mathrm{Data}}{\mathrm{Datarate}\left(m\right)}& \left(9\right)\end{array}$
with t.sub.header being the time duration of the header of the frame, Data being the load in bytes of the PSDU for transmitting in the PPDU frame, Datarate is the instantaneous data rate corresponding to the interface and to the transmission mode m in use.

(28) With reference to FIG. 4, the parameter λ may be estimated by comparing at the access point the number N of frames transmitted over the link during the duration t.sub.OBS: λ=N/t.sub.OBS.

(29) X can be estimated by making use of information contained in the frame headers and more particularly for the PPDU frames: the size of the PSDU and the modulation and coding scheme (MCS) (or the corresponding data rate) in use. Thus,

(30) $\stackrel{\_}{X}=\frac{\underset{1}{\overset{N}{.Math.}}{X}_n}{N}$
with X.sub.n being the duration of the n.sup.th PPDU frame transmitted during the observation duration, and X being the mean size of the N PPDU frames transmitted during the observation duration. The mean access time t.sub.acc to the communication channel during the observation duration is then calculated knowing the values of λ and of X.

(31) For example, IEEE standard 802.11_2012 [2] chapter 20.3.2 PPDU formats distinguishes between three types of PPDU: non-HT, HT-mixed, and HT-greenfield. For those three formats, the header is always constituted by the following parameters: a synchronization field STF, a channel estimation field LTF, and an information field containing the signal SIG. In order to know the size of a PPDU, it is necessary to use the information contained in the signal SIG.

(32) For a non-HT PPDU, the field L-SIG (FIG. 20-5 of the standard) contains the data rate (L_DATARATE) corresponding to the MCS used for coding the PSDU and it contains the size of the PSDU in bytes (L_LENGTH). For a (mixed or greenfield) HT PDDU, the fields HT-SIG.sub.1 and HT-SIG.sub.2 (FIG. 20-6) contain the MCS (MCS), the bandwidth (CH_BANDWIDTH) and the type of guard interval (GI_TYPE) used for transmitting the PSDU, together with its size in bytes (LENGTH). The MCS, the size of the transmission band, and the guard interval make it possible to deduce the equivalent data rate of the PSDU (the data rates are available in Tables 20-30 to 20-40 of the above-mentioned standard). The time duration of the PSDU is obtained by multiplying the data rate by the size of the PSDU.

(33) For example, for a PPDU of HT-mixed type using the mode m defined by MCS=15, GI_TYPE=0 (LONG_GI) and CH_BANDWIDTH=1 (HT_CBW40), Table 2035 of the above-mentioned standard, reproduced in Appendix A, gives an instantaneous data rate D(m) of 270.0 megabits per second (Mbps) with a guard interval of 800 nanoseconds (ns). With a length LENGTH=65,535, the size of the PSDU field to be transmitted is 65,535 bytes and its duration is:

(34) $X_{PSDU}=\frac{65535\times 8}{270.0e^6}\approx 1.9\mathrm{milliseconds}\left(\mathrm{ms}\right)$
The maximum size of the HT-mixed header is 64 microseconds (μs), giving the PPDU frame a size X.sub.m of about 2 ms.

(35) Observing the traffic over the channel during t.sub.OBS thus enables the access point to determine λ and X and to deduce therefrom G and also t.sub.acc and t.sub.tra.sub.m at the instant t that corresponds to the end of the observation period.

(36) The new m.sub.t′ that is to be selected at an instant t′ later than the instant t, t′=t+Δt, must guarantee the same mean data rate as at instant t in order to transmit the service frames n that contain on average Data payload bytes and each of which has a time duration

(37) $t_{{\mathrm{tra}}_{m_{t^{\prime }}}}\text{:}$
D.sub.moy(t,m.sub.t)=D.sub.moy(t′,m.sub.t′)  (10)

(38) A mean data rate takes account of the probability of successfully transmitting the service frames n, i.e. transmitting them without collisions, and it can thus be expressed in the following form:
D.sub.moy(t,m.sub.t)=P.sub.S(G(t,m.sub.t)).Math.λ.sub.n(t,m.sub.t).Math.8.Math.Data  (11)

(39) Given that the transmission channel is shared between the stations and thus between the services required by the users, and given that it has random access, the occupancy G of the channel as given by equation (4) needs to be weighted. This weighting is necessary in order to take account of collisions in order to estimate the traffic that is really transmitted, i.e. without collision: G is thus designated by the incoming traffic, and the traffic without collision S is said to be the outgoing traffic. The expression for outgoing traffic as a function of G depends on the protocol for accessing the channel in use.

(40) For an ALOHA protocol as described in [3], the outgoing traffic S may be expressed in the form:
S=Ge.sup.−2G  (12)

(41) For an ALOHA protocol using time slots, the outgoing traffic S may be expressed in the form:
S=Ge.sup.−G  (13)

(42) With an ALOHA protocol, when the incoming traffic is too great (G.fwdarw.∞), the outgoing traffic is zero since all of the packets are subjected to collisions. In contrast, if the incoming traffic is very low (G.fwdarw.0), then nearly all of the packets are correctly transmitted.

(43) CSMA protocols make it possible to increase the fraction of packets that are correctly transmitted compared with the ALOHA protocol.

(44) For the CSMA/CD as described in Chapter 20 of [2], if the frame retransmission delay is long compared with the duration of the frames, and if the frames are considered to arrive independently, it is known that:

(45) $\begin{array}{cc}S=\frac{G{e}^{-aG}}{G\left(1+2a\right)+e^{-aG}}& \left(14\right)\end{array}$

(46) For CSMA/CA operating with time slots, it is known that:

(47) $\begin{array}{cc}S=\frac{aG{e}^{-aG}}{1+a-e^{-aG}}& \left(15\right)\end{array}$

(48) The probability P.sub.S(G) that a packet will be transmitted successfully without collision over the communication channel while the occupancy of the communication channel is G, is expressed in the following form:

(49) $\begin{array}{cc}{P}_S\left(G\right)=\frac{S}{G}& \left(16\right)\end{array}$

(50) Under certain conditions, with a CSMA protocol, the outgoing data rate can be approximated using the following equation:

(51) $\begin{array}{cc}S\approx \frac{G}{1+G}& \left(17\right)\end{array}$

(52) With a CSMA protocol, the probability P.sub.S(G) that a PPDU frame will be transmitted successfully without collision over the channel can then be written:

(53) $\begin{array}{cc}{P}_S\left(G\right)\approx \frac{1}{1+G}& \left(18\right)\end{array}$

(54) By combining equations (18) and (11), the mean data rate can be expressed in the following form:

(55) $\begin{array}{cc}{D}_{{\mathrm{moy}}_n}\left(t,m_t\right)\approx {\lambda }_n\left(t,m_t\right).Math.\frac{1}{1+G\left(t,m_t\right)}.Math.8.Math.{\mathrm{Data}}_n& \left(19\right)\end{array}$

(56) Let G′ be the fraction of traffic that is due to services other than the service n:
G′=G−λ.sub.nX.sub.n  (20)
which corresponds to a frame arrival frequency:
λ′=λ−λ.sub.n  (21)

(57) The mean data rate can then be expressed in the form:

(58) 0$\begin{array}{cc}{D}_{{\mathrm{moy}}_n}\left(t,m_t\right)\approx \frac{{\lambda }_n\left(t,m_t\right)}{1+G^{\prime }\left(t,m_t\right)+{\lambda }_n\left(t,m_t\right)\stackrel{\_}{X_n}\left(t,m_t\right)}.Math.8.Math.{\mathrm{Data}}_n& \left(22\right)\end{array}$

(59) From equation (22), the transmission rate of service frames n can be expressed in the following form:

(60) $\begin{array}{cc}{\lambda }_n\left(t,m_t\right)\approx \frac{1+G^{\prime }\left(t,m_t\right)}{\frac{8.Math.{\mathrm{Data}}_n}{D_{{\mathrm{moy}}_n}\left(t,m_t\right)}-\stackrel{\_}{X_n}\left(t,m_t\right)}& \left(23\right)\end{array}$

(61) This equation (23) makes it possible to evaluate the successful transmission rate of service frames n at instant t′=1+Δt, as follows:

(62) $\begin{array}{cc}{\lambda }_n\left(t^{\prime },m_{t^{\prime }}\right)\approx \frac{1+G^{\prime }\left(t,m_t\right)}{\frac{8.Math.{\mathrm{Data}}_n}{D_{{\mathrm{moy}}_n}\left(t,m_t\right)}-t_{{\mathrm{tra}}_{{\mathrm{nm}}_{t^{\prime }}}}}& \left(24\right)\end{array}$
assuming that the occupation of the channel due to the services other than the service n remains unchanged at time t′ compared to what it was at time t.

(63) By combining equations (8), (20), and (21), the mean access time

(64) $t_{{\mathrm{acc}}_{{\mathrm{nm}}_{t^{\prime }}}}$
to the communication channel at instant t′ for the frames of the service n that are to be transmitted may be estimated as follows:

(65) $\begin{array}{cc}{t}_{{\mathrm{acc}}_{{\mathrm{nm}}_{t^{\prime }}}}\approx \frac{1-\left(G^{\prime }+{\lambda }_n\left(t^{\prime },m_{t^{\prime }}\right).Math.t_{{\mathrm{tra}}_{{\mathrm{nm}}_{t^{\prime }}}}\right)}{{\lambda }^{\prime }+{\lambda }_n\left(t^{\prime },m_{t^{\prime }}\right)}& \left(25\right)\end{array}$

(66) For a given service n, in order to select at instant t′ a mode m from N modes, the method determines a metric value γ.sub.nm.sub.t for each mode m. This metric amounts to comparing an evaluation of the mean access time

(67) $t_{{\mathrm{tra}}_{{\mathrm{nm}}_{t^{\prime }}}}$
to the communication channel at time t′ with an evaluation of the time

(68) $t_{{\mathrm{acc}}_{{\mathrm{nm}}_{t^{\prime }}}}$
needed at the time t′ for transmitting a PPDU frame relating to the service n with this new mode m:

(69) $\begin{array}{cc}{\gamma }_{{\mathrm{nm}}_{t^{\prime }}}=10.Math.{\log }_{10}\left(\frac{t_{{\mathrm{acc}}_{{\mathrm{nm}}_{t^{\prime }}}}}{t_{{\mathrm{tra}}_{{\mathrm{nm}}_{t^{\prime }}}}}\right)& \left(26\right)\end{array}$

(70) For a mode m, the time

(71) $t_{{\mathrm{acc}}_{{\mathrm{nm}}_{t^{\prime }}}}$
needed for transmitting a frame PPDU is determined from equation (9) and the mean access time

(72) $t_{{\mathrm{acc}}_{{\mathrm{nm}}_{t^{\prime }}}}$
is evaluated from equation (25) by replacing the time

(73) 0$t_{{\mathrm{tra}}_{{\mathrm{nm}}_{t^{\prime }}}}$
by the value determined using equation (9) and replacing λ.sub.n(t′,m.sub.t′) by the value determined using equation (24). Knowing the values of

(74) $t_{{\mathrm{tra}}_{{\mathrm{nm}}_{t^{\prime }}}}$
and of

(75) $t_{{\mathrm{tra}}_{{\mathrm{nm}}_{t^{\prime }}}}$
the method determines the metric γ.sub.nm.sub.t′.

(76) The method compares the value of the metric γ.sub.nm.sub.t′ with a threshold of value γ.sub.min that can be deduced by simulations. If the metric γ.sub.nm.sub.t′ is below the threshold, then the mode m is eliminated from the selection of the link adaptation method.

(77) The minimum value γ.sub.min of the threshold is zero, which corresponds to times

(78) $t_{{\mathrm{tra}}_{{\mathrm{nm}}_{t^{\prime }}}}\mathrm{and}{t}_{{\mathrm{acc}}_{{\mathrm{nm}}_{t^{\prime }}}}$
that are practically identical, with channel occupancy being at a maximum; there are no longer any “holes” between the frames that are transmitted. The condition γ.sub.min=0 requires the mean access time to the channel for a service n to be equal to not less than the time for transmitting a frame of the service n.

(79) Once the mode m for the station n has been selected, the occupancy G of the link and the overall arrival rate of frames λ can be approximated by the following equations:
G≈G′+G.sub.nm.sub.t′  (27)
λ≈λ′+λ.sub.n(t′,m.sub.t′)  (28)

(80) In a first implementation of the invention, preselection is performed with the help of the metric γ from among the various transmission modes Mode.sub.1, . . . , Mode.sub.M of the entity corresponding to a service n requested by a user of the entity. This preselection serves to determine which mode can be used for performing transmission while taking account of the state of the transmission channel. In other words, the preselection serves to put aside modes that cannot enable PPDU frames to be transmitted while guaranteeing a mean data rate D.sub.moy, a quality of service QoS, and optimum occupation of the channel.

(81) If a plurality of transmission modes are available after this preselection, selection is performed on the preselected modes by using another metric.

(82) FIG. 5 is a flow chart of the main steps of performing the method in a first implementation of the invention in a single technology context, i.e. all of the modes considered belong to a single given transmission interface.

(83) In the mode described, the access point has various transmission modes that are associated with respective different instantaneous data rates and that perform the method 1 of the invention. In the mode shown, a mode counter is initialized at one, m=1, and the set A of preselected modes is initialized as an empty set.

(84) In a first step 2, the method calculates the metric γ.sub.n,m for the service n and the mode m over an observation window t.sub.OBS.

(85) In a second step 3, the method compares the metric γ.sub.n,m with a threshold value γ.sub.n,m: γ.sub.n,m>γ.sub.n,m?

(86) If the metric is strictly greater than the threshold, then the mode m is preselected, i.e. the mode Mode.sub.m belongs to the set A: Mode.sub.mεA, and the method then performs a test to determine whether the value of the counter is less than M: m≦M?, i.e. whether any more modes remain.

(87) Else, the method performs the test.

(88) If during the test the value of the counter is less than M, then the counter is incremented by one, m=m+1, and the method loops back to the first step.

(89) Else, during a third step 4, the method then makes a selection using a determined criterion in order to select one of the transmission modes from the various preselected modes that belong to the set A.

(90) This selection may be performed, by way of example, by implementing a known link adaptation algorithm.

(91) In a first implementation shown in FIG. 6, the link adaptation algorithm makes use of the green link budget (GLB) metric {α,β} constituted by two parameters α and β using the method described in patent application WO 2011/083238 published on Jul. 14, 2011. FIG. 6 shows only the third step of FIG. 5.

(92) This mode is particularly suitable when the transmission modes are associated with different transmission interfaces. This GLB {α,β} metric serves to select the transmission interface and the transmission mode that makes it possible to minimize the transmission power while complying with an instantaneous data rate constraint, a QoS constraint, and an emitter-receiver distance. It is constituted by two parameters α and β.

(93) The parameter α serves to select the transmission Interface and the transmission mode subjected to the least multi-path degradation in the physical layer. It is the sum of two sub-parameters:
α=MCM+PLM
where MCM is the multi-path channel margin that corresponds to the power difference required between the ideal circumstances and realistic transmission circumstances:
MCM=SNR.sub.c−SNR.sub.c.sup.AWGN
MCM=S.sub.m−S.sub.0
where SNR.sub.c and SNR.sub.c.sup.AWGN are respectively the signal-to-noise ratios (SNRs) required in the realistic situation and in the ideal situation (AWGN channel). S.sub.m and S.sub.0 are the respective powers required in the realistic situation and in the ideal situation (AWGN channel).

(94) Where PLM is the relative propagation attenuation that corresponds to the difference of losses due to propagation between the ideal situation and the realistic transmission situation:

(95) $\mathrm{PLM}=\mathrm{PL}\left(d,f_c\right)-{\mathrm{PL}}_{\mathrm{FS}}\left(d,f_c\right)$$\mathrm{PLM}=10.Math.n.Math.{\log }_{10}\left(\frac{d}{d_0}\right)-10.Math.{{\log }_{10}\left(\frac{d}{d_0}\right)}^2+\sigma$$\mathrm{PLM}=10.Math.{{\log }_{10}\left(\frac{d}{d_0}\right)}^{n-2}+\sigma$
where PL(d) represents propagation losses in the realistic situation (multiple paths), PL.sub.FS(d) represents propagation losses in ideal circumstances (free space), d is the distance to the target, f.sub.c is the carrier frequency, d.sub.0 is the reference distance, n characterizes the multi-path channel, and corresponds to the exponential decrease due to distance, and σ corresponds to masking effects.

(96) In step 5, the method calculates the metric α using the equation:
α=(S.sub.M−S.sub.0)+(PL(d,ƒ.sub.c)−PL.sub.FS(d,ƒ.sub.c)).

(97) The parameter β corresponds to the difference between the power available and the power required for ensuring a QoS (i.e. a target BER evaluated over the MSDU data field) with an emitter-receiver distance d. This parameter makes it possible to adjust the transmission power of the selected transmission mode interface in order to reduce the power that is radiated and in order to control power dynamically.
β=ARP(d,ƒ.sub.c)−S.sub.M

(98) The mean receive power ARP(d) is expressed as follows:
ARP(d,ƒ.sub.c)=EIRP+G.sub.RPL(d,ƒ.sub.c)
where EIRP is the equivalent isotropic radiated power from the emitter, and G.sub.R is the gain of the receive antenna.

(99) The method calculates in step 5 the metric β which is expressed using the following equation:
β=EIRP+G.sub.R−α−S.sub.0−PL.sub.FS(d,ƒ.sub.c)  (30)

(100) Thus, the GLB {α,β} algorithm performs a selection in step 6 to select a transmission mode in a multimode terminal by optimizing the normalized link budget that makes it possible to compare systems operating in different bands (RF, optical, wired) and that present independent power variation ranges. The metric α evaluates the overall relative degradation introduced by the multi-path channel on the transmission device and in the propagation medium. The metric μ measures the power excess between the power available ARP(d,f.sub.c) and the power required S.sub.M for providing communication with a given QoS. Both of these metrics are measured in decibels (dB).

(101) In a second implementation shown in FIG. 7, the link adaptation algorithm is an algorithm of the MiSer type. FIG. 7 shows only the third step of FIG. 5. The MiSer algorithm comprises an initialization stage 7 during which quality tables are calculated and it also comprises a real time execution stage.

(102) During the initialization stage 7, the method estimates two parameters:

(103) L(R,P.sub.T,l,s,SRC,LRC): the number of bits that will be transmitted correctly; and

(104) E(R,P.sub.T,l,s,SRC,LRC): the energy needed to transmit L.

(105) These two estimates are calculated recursively by considering all possible combinations of the parameters R, P.sub.T, l, s, SRC, and LRC.

(106) Thereafter, from the estimates of the parameters, the method estimates the energy efficiency J in application of equation (1) during the real time execution stage, by considering the following six parameters: R, P.sub.T, l, s, SRC, and LRC. The station that seeks to transmit a load £ estimates in step 8 the propagation losses s and recovers the number of attempts at transmitting the RTS (SRC) and at transmitting the PPDU (LRC). As a function of the data quadruplet (l,s,SRC,LRC), the station retrieves from the tables that were calculated while it was not in operation, the combination of parameters (R,P.sub.T) that is to be used in order to maximize the energy efficiency of the transmission. For each of the possible combinations (l,s,SRC,LRC), the combination (R,P.sub.T) that minimizes J is stored.

(107) In step 9, the method selects the mode m that maximizes the energy efficiency of transmission.

(108) This algorithm thus makes it possible to maximize the efficiency of transmitting a load of data but it is quite constraining since it requires use to be made of RTS/CTS (thereby increasing the consumption of the system since emitting RTS/CTS frames has an energy cost). Furthermore, it is necessary while not in operation to calculate the energy efficiencies that correspond to all possible combinations of the parameters (R,P.sub.T,l,s,SRC,LRC).

(109) Using RTS/CTS frames makes it possible to reduce the number of collisions due to the hidden station problem, but it encumbers the channel (for each PSDU that is to be transmitted, it is necessary beforehand to emit an RTS and to receive a CTS). This type of MiSer algorithm advantageously combines optimizing the modulation and coding scheme (MCS) with managing the transmission power.

(110) In a variant, the MiSer algorithm is adapted to be performed without activating the RTS/CTS mechanism. In this adaptation, the algorithm estimates the energy efficiency using the following equation:

(111) $\begin{array}{cc}{\eta }_{\mathrm{nm}}^p=\frac{L_{\mathrm{nm}}\text{:}\mathrm{quantity}\mathrm{of}\mathrm{bits}\mathrm{transmitted}\mathrm{successfully}}{E_{\mathrm{nm}}\text{:}\mathrm{energy}\mathrm{expended}\mathrm{for}\mathrm{transmission}}& \left(31\right)\end{array}$

(112) For this purpose, the method estimates the following parameters:
L.sub.nm=P.sub.S(G.sub.nm).Math.8.Math.Data.sub.n
E.sub.nm=t.sub.tra.sub.nm.Math.P.sub.T.sub.nm  (32)
where P.sub.T.sub.nm is the power needed on emitting to address each of the stations with the desired quality of service, e.g. with a BER=10.sup.−5 (with a safety margin of 3 dB). This power is expressed in the following form:
P.sub.T.sub.nm=EIRP+G.sub.R−β.sub.nm+3 dB  (33)
where EIRP is the equivalent isotropic radiated power of the emitter, and G.sub.R is the gain of the receive antenna.

(113) The method estimates the energy efficiency at instant t for a service n:

(114) $\begin{array}{cc}{\eta }_{{\mathrm{nm}}_t}=\frac{8.Math.{\mathrm{Data}}_n}{\left(1+G^{\prime }+{\lambda }_n\left(t,m_t\right)t_{{\mathrm{tra}}_{\mathrm{nm}}}\right).Math.t_{{\mathrm{tra}}_{{\mathrm{nm}}_{t^{\prime }}}}.Math.P_{T_n}}& \left(34\right)\end{array}$

(115) In order to determine the mode m.sub.t′ that is to be selected at instant t′ for the service n, the method estimates the forecast energy efficiency at that instant t′:

(116) $\begin{array}{cc}{\eta }_{{\mathrm{nm}}_t}=\frac{8.Math.{\mathrm{Data}}_n}{\left(1+G^{\prime }+{\lambda }_n\left(t^{\prime },m_t\right)t_{{\mathrm{tra}}_{{\mathrm{nm}}_{t^{\prime }}}}\right).Math.t_{{\mathrm{tra}}_{{\mathrm{nm}}_{t^{\prime }}}}.Math.P_{T_n}}& \left(35\right)\end{array}$

(117) Selection using the adaptive MiSer algorithm consists in selecting the mode that optimizes η.sub.nm.sub.t′.

(118) FIG. 8 shows an implementation in a context of adapting a single technology link. In this example, the WiFi type communication system has an access point AP and two stations STA.sub.1, STA.sub.2.

(119) The station STA.sub.1 is spaced apart from the access point AP by a distance d.sub.1=4 meters (m) and it seeks to set up communication with the AP at a mean data rate D.sub.1=7.5 megabits per second (Mbps). The station STA.sub.2 is spaced apart from the access point AP by a distance d.sub.2=6 m and it seeks to set up communication with the AP at a mean data rate D.sub.2=7.5 Mbps. Power control is performed so that each transmission takes place with the minimum required power. The two stations have the following parameters: carrier frequency: f.sub.c=5 GHz; transmission bandwidth: B=40 megahertz (MHz); number of spatial streams per station: N.sub.SS=2; maximum radiated power: EIRP=23 dBm; antenna receive gain: G.sub.R=0 dB; mean payload of PSDUs: data=8000 bytes; duration of headers: t.sub.header=20 μs.

(120) The method of accessing the channel is of the CSMA/CA type and the access point estimates the following parameters at instant t: λ, λ.sub.n, X, X.sub.n. The access point and each of the stations has only one transmission interface of 802.11 type and has the following ten transmission modes:

(121) TABLE-US-00001 m 1 2 3 4 5 6 7 8 9 10 MIMO SDM STBC MCS QPSK ½ QPSK ¾ 16-QAM ½ 16-QAM ¾ 64-QAM ⅔ QPSK ½ QPSK ¾ 16-QAM ½ 16-QAM ¾ 64-QAM ⅔
The associated instantaneous data rates D.sub.m are as follows:

(122) TABLE-US-00002 m 1 2 3 4 5 6 7 8 9 10 D.sub.m (Mbps) 54 81 108 162 216 27 40.5 54 81 108
At time t=0, the AP uses mode 6 to deliver a mean data rate D.sub.1=7.5 Mbps to the station STA.sub.1. Link occupancy is then G=38.96%, λ.sub.1=163 frames/s.

(123) In compliance with the description of an implementation of the method of the invention as shown in FIG. 5, the time t.sub.tra.sub.nm is determined from equation (9) for each of the modes and for the service 2 associated with the station 2:

(124) TABLE-US-00003 m 1 2 3 4 5 6 7 8 9 10 t.sub.tra.sub.2m 1.21 0.81 0.61 0.42 0.32 2.39 1.60 1.21 0.81 0.61
The method evaluates the time t.sub.acc.sub.2m and calculates the metric γ.sub.2m with the constraint of guaranteeing the data rate D.sub.2=75 Mbps to the station STA.sub.2:

(125) TABLE-US-00004 m 1 2 3 4 5 6 7 8 9 10 t.sub.acc.sub.2m 1.08 1.35 1.49 1.61 1.68 0.18 0.80 1.08 1.35 1.49 γ.sub.2m −0.47 2.23 3.85 5.90 7.24 −11.26 −3.03 −0.47 2.23 3.85
The metric γ.sub.2m makes it possible to eliminate the modes 1, 6, 7, and 8 from the selection since the values obtained for this metric are below the minimum threshold γ.sub.min=0.

(126) Selection is then performed using a link adaptation algorithm.

(127) In a first implementation, the method of the invention performs a GLB algorithm in compliance with the implementation described with reference to FIG. 6. The parameters S.sub.m and S.sub.0 are given in [4] by considering a propagation channel of TGn model B type. The parameters PLd, 5 GHz) and PL.sub.fs(D, 5 GHz) are shown in FIG. 9 showing propagation loss models. The method calculates:

(128) TABLE-US-00005 m 1 2 3 4 5 6 7 8 9 10 α.sub.2m 15.08 18.12 17.12 17.76 19.11 7.08 8.12 8.12 7.16 8.61 β.sub.2m 21.78 16.78 14.78 10.18 4.78 29.78 26.78 23.78 20.78 15.28

(129) The metric α leads to selecting mode 9. Thus, the method of the invention is advantageous since it makes it possible to improve the efficiency of the GLB algorithm. The GLB algorithm would lead to selecting mode 6. In the invention, the load on the channel is evaluated by means of the metric γ.sub.2m. The method of the invention makes it possible to avoid overloading the channel by eliminating selection of mode 6 which would lead to the channel being overloaded.

(130) In a second implementation, the method of the invention performs a MiSer algorithm adapted in compliance with the implementation described with reference to FIG. 7.

(131) Using equation (34), the method initially calculates η.sub.1=171.13 Gbit/J. Using equation (35) the method then calculates:

(132) TABLE-US-00006 m 1 2 3 4 5 6 7 8 9 10 η.sub.2m.sub.t.sub.. 12.41 6.15 5.27 2.76 1.06 33.10 27.97 19.67 15.46 5.91

(133) Determining the energy efficiency η.sub.2m.sub.t′ then leads to selecting mode 9. Thus, the method of the invention is advantageous since it makes it possible to improve the efficiency of the adapted MiSer algorithm. The adapted MiSer algorithm would lead to selecting mode 6. With the invention, the load on the channel is evaluated by means of the metric γ.sub.2m. The method of the invention makes it possible to avoid overloading the channel by eliminating selection of mode 6 which leads to the channel being overloaded.

(134) The simplified structure of an access point performing a link adaptation method of the invention is described with reference to FIG. 10.

(135) Such an access point AP comprises a memory module 100 having a buffer random access memory (RAM), a processor unit 101, e.g. having a microprocessor μP that is controlled by the computer program 102 that performs the link adaptation method of the invention.

(136) On initialization, the code instructions of the computer program 62 are loaded by way of example into the RAM prior to being executed by the processor of the processor unit 61. The processor unit 61 receives as input the data transmitted over the channel. The microprocessor of the processor unit 61 performs the steps of the above-described link adaptation method in compliance with the instructions of the computer program 62 in order to determine the time occupancy metric γ of the channel.

(137) For this purpose, the access point comprises: M transmission modes associated respectively with different instantaneous data rates; a determination module for determining the value of a first time occupancy metric γ.sub.n,m of the channel for a given transmission mode; a comparator module for comparing the values of the metric γ.sub.n,m for different modes with a threshold value in order to preselect transmission modes; and a link adaptation module for making a selection in application of a determined criterion from among the various modes that have been preselected.

(138) While the processor of the processor unit 61 is executing the code instructions of the computer program 62: the processor initializes a mode counter at one and a variable containing the set A of preselected modes to an empty set; the determination module calculates the metric γ.sub.n,m for the service n and for the mode m over an observation window t.sub.OBS; the processor compares the metric γ.sub.n,m with a threshold value for γ.sub.n,m: γ.sub.n,m>γ.sub.n,m? If the metric is strictly greater than the threshold, then the mode m is preselected, i.e. the mode Mode.sub.m is stored in the variable containing the set A: Mode.sub.mεA. The processor compares the value of the counter with M: m≦M?, i.e. it determines whether there remain any more modes. If the value of the counter is less than M during the test, then the mode counter is incremented by one, m=m+1; and the link adaptation module makes its selection in application of a determined criterion from among the various preselected modes that belong to the set A once the value of the counter becomes strictly greater than M.

(139) TABLE-US-00007 TABLE 20-35 MCS parameters for optional 40 MHz, NSS = 2, NES = 1 MCS Data rate (Mb/s) Index Modulation R N.sub.BPSCS(i.sub.SS) N.sub.SD N.sub.SP N.sub.CBPS N.sub.DSPS 800 ns GI 400 ns GI 8 BPSK ½ 1 108 6 216 108 27.0 30.0 9 QPSK ½ 2 108 6 432 216 54.0 60.0 10 QPSK ¾ 2 108 6 432 324 81.0 90.0 11 16-QAM ½ 4 108 6 864 432 108.0 120.0 12 16-QAM ¾ 4 108 6 864 648 162.0 180.0 13 64-QAM ⅔ 6 108 6 1296 864 216.0 240.0 14 64-QAM ¾ 6 108 6 1296 972 243.0 270.0 15 64-QAM ⅚ 6 108 6 1296 1080 270.0 300.0

REFERENCES

(140) [1] Daji Qiao, Sunghyun Choi, Amit Jain, and Kan G. Shin, “MiSer: An Optimal Low-Energy Transmission Strategy for IEEE 802.11a/h,” MobiCom '03, pp. 161-175, September 2003. [2] IEEE Computer Society, “Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specifications,” IEEE Std 802.11™-2012, 2012 [3] Chapter 16, Section 5, pp. 1068-1077, John G. Proakis and Masoud Salehi, Digital. Communications, 5th ed., publisher McGraw-Hill International, 2008. [4] V. Erceg et al., “TGn Channel Models,” IEEE P802.11 Wireless LANs, 2007.

(141) Although the present disclosure has been described with reference to one or more examples, workers skilled in the art will recognize that changes may be made in form and detail without departing from the scope of the disclosure and/or the appended claims.