FR2965696A1 - CHANNEL SELECTION METHOD BY TRANSMITTER, DATA TRANSMITTING METHOD AND DEVICE, AND COMPUTER PROGRAM - Google Patents

Abstract

A method of selecting, by a transmitter, a transmission channel from among a plurality of transmission channels through which the transmitter can transmit data to a receiver, comprises the steps of: - determining (60, 70, 80) a distribution of probabilities of using a channel of the plurality as a function of at least one value of a measured transmitter performance metric in at least one of the plurality; - selection (90) of a channel by random draw according to the distribution of probabilities of determined use. A method and an apparatus for transmitting data and an associated computer program are also described.

Description

The invention relates to a method of channel selection by a transmitter, and to a method and an apparatus for transmitting data and an associated computer program. In some communication systems between a transmitting device (transmitter) and a receiving device (receiver), a given number of transmission channels (or communication channels) are available to enable the exchange of data of a given transmitter to the associated receiver. This is particularly the case for Wi-Fi® wireless communication systems or in general for systems based on technologies such as OFDMA or multi-carrier CDMA. Different solutions have been proposed for assigning a given transmission channel to the communication between a transmitter and a receiver.

For example, it has been planned that a given equipment (transmitter) will always transmit in the same channel (chosen once and for all by the equipment manufacturer or by the equipment operator when it is put into service by example). It has also been envisaged that the selection of the channel used by a transmitter to transmit data is made at the beginning of each communication established by the transmitter. The choice of the channel can then be made in different ways, for example according to a quality criterion of the communication (such as the optimization of a signal-to-noise ratio plus RSBI interference) or arbitrarily and randomly following typically a uniform probability law. It has also been found that, in such systems, the transmission rate is strongly affected by the presence of several devices close to each other. In this case, the transmitters all have the possibility of transmitting on the same set of channels and use a particular channel determined by each transmitter according to the methods mentioned above. Transmitters therefore seldom make optimal use of the available channels in the absence of a coordinating entity that would best govern the allocation of channels.

The aim of the invention is to improve this state of affairs and thus proposes a method of selecting, by a transmitter, a transmission channel from among a plurality of transmission channels through which the transmitter can transmit data to a receiver, characterized by the following steps: - determining a distribution of probabilities of use of a channel of the plurality as a function of at least one value of a transmitter performance metric measured in at least one channel of the plurality ; - Selection of a channel by random draw according to the distribution of probabilities of determined use.

This combines the flexibility of a random choice (since any channel is potentially selected by such a random draw) and the consideration of the transmitter's performance in at least one channel (which makes it possible to direct the choice towards the or the channels in which the performance metric indicates good results).

The determination and selection steps can be repeated during transmission of a data set by the transmitter, which allows dynamic selection of the channel used for transmission. It can then be predicted, for example, that the probability of using a given channel for the current iteration is determined as a function of the difference between an average value of the performance metric, measured in the channel given at each iteration between a first iteration and a second iteration (not necessarily consecutive) prior to the current iteration, and an average value of the performance metric, measured in the selected channels between the first iteration and the second iteration. This technique makes it possible to compare, over a given period (between the first iteration and the second iteration), the performance obtained on the channel given to the performance actually obtained due to the dynamic channel change. The channels to be favored for subsequent emissions can be deduced (by defining the distribution of probabilities of use so that these channels have a greater probability of being used). The performance metric is for example defined according to a measurable parameter at the receiver (such as multiple access interference) and / or a set of measurable parameters at the transmitter (for example its power of transmission), as in the example described below. The use of such parameters may allow, as will emerge from the example described, to calculate the performance metric even on channels that have not been used, and thus to calculate, for each channel, a value evaluating the regret not having used the channel concerned. We note also that the use of previous iterations allows learning and convergence to the most favorable channels. Precisely, the use of the previous iterations allows an apprenticeship of the optimal joint distribution of channels, such that, once this distribution is reached, none of the emitters would be interested in using a different probability distribution; it would indeed get a lower average performance. In game theory, this optimal distribution is known as "correlated equilibrium". For example, it can be provided that the second iteration is the iteration preceding the current iteration, which allows the latest measurements to be taken into account. It can also be expected that the first iteration corresponds to the beginning of a new communication. The determination and selection steps can also be repeated periodically. This may be a repetition based on a period of a fixed fixed duration, or a repetition for each packet of data to be transmitted, as is the case in the example presented below. . The distribution of probabilities of use is for example determined according to a plurality of values obtained by measuring at least one performance metric of the transmitter in each channel of the plurality of channels. We are here targeting typically all accessible channels. The use of values relating to several or all the available channels makes it possible to coherently define a probability distribution relative to all the accessible channels and thus to obtain an appropriate selection in all the channels.

The value of the performance metric may in practice be a function of at least one measurement made by a receiver. The performance metric uses, for example, at least one of the following metrics: signal-to-noise ratio plus interference, multiple access interference level, attainable maximum transmission rate (or Shannon rate), correctly transmitted packet rate. These types of metrics are currently available and thus allow the invention to be implemented within existing systems. The invention also proposes a method for transmitting data by a transmitter, characterized in that it comprises a step of selecting a channel in accordance with what has just been explained and a data transmission step (for example a packet of data) to the receiver in the selected channel.

The invention likewise proposes a transmission device able to transmit data (or information) intended for a receiver in a plurality of transmission channels, characterized in that it comprises: means for determining a distribution of probabilities of using one of the plurality as a function of at least one value of a transmitter performance metric in at least one of the plurality; means for selection of a channel by random draw according to the distribution of probabilities of determined use. As already indicated in connection with the method, the determination means and the selection means may be designed to implement a plurality of iterations comprising the determination of a probability distribution and the selection of a channel. Likewise, it can be provided that the means for determining the probability distribution are able to determine the probability of using a given channel for the current iteration as a function of the difference between an average value of the performance metric. , measured in the given channel at each iteration between a first iteration and a second iteration prior to the current iteration, and an average value of the performance metric, measured in the selected channels between the first iteration and the second iteration. The determining means may determine the usage probability distribution based on a plurality of values obtained by measuring at least one transmitter performance metric in each channel of the plurality of channels. It is also possible to provide means for determining the value of the performance metric as a function of at least one measurement performed by a receiver. Finally, a computer program is proposed comprising instructions for implementing the method presented above when this program is executed by a processor. The optional features and associated advantages, expressed above with respect to the method, may also apply to the device and computer program just mentioned.

Other characteristics and advantages of the invention will emerge in the light of the description which follows, given with reference to the appended drawings, in which: FIG. 1 schematically represents an exemplary context of implementation of the invention; FIG. 2 represents the main steps of a channel selection and data exchange method performed in accordance with the teachings of the invention; - Figure 3 shows the main elements of an example transmitter. FIG. 1 represents an example of context in which the invention can be implemented. Each emitter Ek among a plurality of emitters E ,, E2,..., EK exchanges data with a receiver Ri from a plurality of receivers R1, R2,..., Rj (generally by means of the generation by the emitter Ek of electrical or electromagnetic signals representative of the data and transmitted to the receiver Ri). To do this, the transmitter has at its disposal a plurality of communication channels (or transmission channels) C ,, C2, ..., Cs through which it can transmit the data to the recipient receiver. These channels each correspond for example to a usable frequency band for communication between the transmitter and the receiver (in which case the communication is carried out by transmitting from the transmitter to the receiver electrical or electromagnetic signals in the frequency band concerned). The communications between the transmitter and the receiver can be carried out by means of radio waves (so-called "wireless" communications), for example of the Wi-Fi® type (in which case the communication channels have properties defined in the IEEE 802.11 standard). The invention however also applies to wired communications. Note that it is not expected in the situation presented in FIG. 1 of a higher regulatory entity that would organize the distribution and allocation of the different channels to the different communications established between a transmitter and a receiver. On the contrary, each transmitter independently selects the channel in which it transmits data (that is, in practice electrical or electromagnetic signals representative of this data) according to a process explained hereinafter.

Although a receiver associated with each transmitter has been shown in FIG. 1, it is conceivable in practice that the same receiver receives signals transmitted by several transmitters (in this case possibly on different channels selected by each transmitter according to the process explained in the following).

Furthermore, the invention is explained here in the case of an example where a single receiver is associated with each transmitter. As a variant, it would be possible for several receivers to receive the data transmitted by a single transmitter. FIG. 2 represents the main steps of a method that can be envisaged for the selection and transmission of data by an emitter Ek. It is considered here that the data are organized in packets to be emitted successively by the transmitter Ek. We note that the equalities that appear on most of the steps shown in Figure 2 are intended to provide an overview of the functionality of the step concerned but do not constitute mathematical equations (which are given below).

We are therefore interested in the following to a single emitter pair Ek - receiver Ri. In the embodiment described here, it was chosen for decision making at time t to rely on the performance metric values between a first iteration corresponding to the beginning of communication and a second iteration corresponding to the first one. iteration immediately preceding the current iteration. It could alternatively be interested only in a part of these iterations, for example at each moment to a predetermined number (for example equal to 10) of iterations that precede the current iteration. The process starts at a step 10 at which a time variable t is initialized to 0. This step corresponds to the beginning of the period during which the probability distribution used for the channel selection will be defined as indicated below.

Thus, the probability distribution between the different channels (potentially used by the transmitter) is also initialized to a uniform DISTRo distribution during step 20: for any channel s, the initial probability gk, s (0) is fixed at 1 / S. In addition, a regret value rk, s (0) associated with each channel s (regret value whose role will be explained hereinafter) is initialized, for example to the value 1 (or alternatively to another non-zero value). identical for all channels). We then proceed (step 30) to the selection of a channel sk (0) (for channel selected by the transmitter Ek at time t = 0, a step denoted So in FIG. 2) by random draw according to a law of probability according to the distribution defined by the variables gk, s (0) (s varying from 1 to S), or at this stage a uniform distribution. In the present embodiment, it is assumed that the transmitters do not have prior experience, so that at this initial stage the different channels have an identical probability of being selected. Step 40 consists for the transmitter Ek to send a data packet to the receiver Ri through the selected channel sk (0).

The step of initializing the process (steps 10 to 40) has been described and steps (50 to 100) implemented at each subsequent instant (PO) are now described. Firstly, the time variable t is incremented at step 50 (which corresponds in the example given here to the processing of the next packet).

Then, in step 60, maximum theoretical rates uk, s (t) are determined, each relating to a channel s as a function of measurements made in the channel s concerned during the transmission of the previous instant (t-1), here when transmitting the previous packet. As already indicated, these magnitudes were noted schematically in FIG. 2 (DEBt representing the theoretical flow rates at the current instant and MESm the measurements at the previous instant). These measurements are, for example, the interference values of multiple access and noise Si, s (t-1) each relating to a channel s and measured by the receiver Ri during the previous transmission. This can be found in Chapter 2 of Fundamentals of Wireless Communication by D. Tse and P. Viswanath, Cambridge University Press, 2005. The use of alternative performance metrics is alternative. If we denote hks ~ (t) the channel gain between a transmitter Ek and a receiver R) on the channel s at time t, the value of multiple access interference and noise measured by the receiver R) is written : hksj (t) K js (t) - 6s + E Pk, max k = 1 2 1 {sk (t) = s}, where 1 {if, (t) = s} is the indicator function equal to 1 when sk (t) = s and equal to 0 otherwise, Pk, max is the maximum transmission power of the transmitter Ek and 6S represents the noise and sk (t) represents the channel chosen by the transmitter Ek at time t. It is proposed here to use as the theoretical maximum flow rate Shannon Pk, max I / lk (t-1) 12 corresponding: uk, s (t) = log (1 + Tj> s (t-1) -Pk, .. hk j (t-1) Z). Let uk (t) be the effective flow at time t in the channel sk (t). In general, we can write uk (t) as:

z s Pk, max I hksj (t -1) 1. 1 {sk (, I) = s1 uk (t) -E log z (1 + i z 7). s-1 0k, s (t-1) - Pk, max hk j (t -1) 1. 1 {sx (! - 1) = s} Indeed, in this sum, only the term relative to the index s such that sk (t-1) = s is non-zero. The rates thus determined make it possible to evaluate for each channel s the regret rk, s (t) of the transmitter Ek for not having used this channel at all the iterations taken into consideration (step 70): this regret is calculated in comparing (by difference) the theoretical maximum flow that would have been obtained by using the channel concerned repeatedly, and the theoretical maximum flow rate obtained on the successively used channels: t - 1 t - 1 rk, s (t) - 1 - E Uk, s (n) 1 2 - l Uk (n) t- 2 n = 2 t n-2 We then define in step 80 the (new) distribution of probabilities DISTRt between the different channels taking into account the regrets evaluated (noted REGt in Figure 2), for example by giving, to the only channels where the regret is sufficiently large (that is to say whose repeated use would have improved the situation), a weighting proportional to the value of the regret : for each channel s, the probability gk, s (t) assigned to this channel is then written max (n , rk, s (t »qk s (t) = s (where max (a, b) is equal to the highest value among a and b, where

E max (z, rk, '(t' n = 1

is a positive or zero value that determines the minimum probability assigned to each channel and where the denominator ensures that the sum of the different probabilities is equal to 1). We can take for example n = 0, in which case the probability of using certain channels may be zero at a given instant t (when the regret to use this channel is negative or zero). Alternatively, one can take a strictly positive value for n so that there will always be a non-zero probability of using each channel.

The selection of the channel sk (t) to be used by the transmitter (step noted S, in FIG. 2) to transmit at the instant (or iteration) t (here to transmit the corresponding packet) can thus be carried out at the step 90 by random draw (or determination) among the available S channels according to the channel probability distribution determined in step 80. This means that a channel has a probability equal to gk, s (t) of being selected at the iteration t.

It is understood that this probability being proportional to the regret of not having used the channel concerned, the transmitter will use after a learning phase the channel (or channels) which is (are) the most favorable (s). This can be seen in the article "A Simple Adaptive Procedure Leading to Correlated Equilibrium" by Hart and Mas-Colell, in Econometrica 68 (2000), 5, 1127-1150.

The transmitter then proceeds to step 100 with the data transmission (here of a data packet) using the selected channel sk (t).

Once these data are sent, the method loops back to step 50 for a new iteration. It is also possible to interrupt the process under certain conditions, for example when all the data (here all the data packets to be sent) have been processed. However, the method will resume when new data are to be transmitted, preferably where it has been interrupted (with a strictly positive value of t) in order to benefit from the learning previously carried out. As a variant, it is possible to provide for the method to resume in its entirety (in particular the initialization steps 10 to 40) when new data are to be transmitted, in particular in the case where the environment (where the channels are located) is quickly changing. FIG. 3 shows the basic elements of an exemplary transmitter in which the method just described is implemented.

This transmitter comprises a processor PROC, for example a microprocessor, which implements the steps of the method just described (in particular the selection of a channel for the transmission of the data) because of the execution of instructions within it. These instructions, which form a computer program, are stored in an MEM memory connected to the microprocessor.

The memory MEM also stores the values of the various variables used during the implementation of the method, namely for example the data to be sent (that is to say, to be transmitted), the maximum theoretical flow values uk, s (t), of regret rk, s (t) and probability gk, s (t). The transmitter also comprises an emitting module EMIS which generates, as functions of commands of the processor PROC, electrical signals (transmitted by an antenna ANT) representing the data to be transmitted, which causes the emission of electromagnetic signals representative of these data in the channel sk (t) selected by the method implemented by the processor PROC. The transmitter furthermore comprises a reception module REC which is also connected to the antenna ANT and which receives in particular the measured values of performance at the receiver interacting with the transmitter, here the values of multiple access interference and of noise (k, s (t), and communicates them to the processor PROC The preceding embodiments are only possible examples of implementation of the invention, which is not limited thereto.

Claims (15)

REVENDICATIONS1. A method of selecting, by a transmitter (E1; E2; EK), a transmission channel from a plurality of transmission channels (CI; C2; C3; CS) through which the transmitter can transmit data to a receiver (R 1, R 2, R 1), characterized by the following steps: determining (60, 70, 80) a distribution of probabilities of using a channel of the plurality as a function of at least one value of a transmitter performance metric measured in at least one channel of the plurality; - selection (90) of a channel by random draw according to the distribution of probabilities of determined use. 2. Selection method according to claim 1, characterized in that the determination and selection steps are repeated during the transmission of a set of data by the transmitter. A selection method according to claim 2, characterized in that the probability of using a given channel at the current iteration is determined as a function of the difference between an average value of the performance metric, measured in the channel given at each iteration between a first iteration and a second iteration prior to the current iteration, and an average value of the performance metric, measured in the selected channels between the first iteration and the second iteration. 4. A selection method according to claim 3, characterized in that the second iteration is the iteration preceding the current iteration. 5. Selection process according to one of claims 1 to 4, characterized in that the determination and selection steps are repeated periodically. 6. A selection method according to one of claims 1 to 5, characterized in that the distribution of probabilities of use is determined according to a plurality of values obtained by performing a measurement of at least one performance metric of the transmitter in each channel of the plurality of channels. 7. Selection method according to one of claims 1 to 6, characterized in that the value of the performance metric is a function of at least one measurement performed by a receiver. 35 8. Selection method according to one of claims 1 to 7, characterized in that the performance metric uses at least one of the following metrics: signal-to-noise ratio plus interference, multiple access interference level, bit rate Maximum achievable transmission rate, correctly transmitted packet rate. 9. A method of transmitting data by a transmitter characterized in that it comprises a channel selection step according to a method according to one of claims 1 to 8, and a step of transmitting data to the receiver in the selected channel. Transmitting device capable of transmitting data intended for a receiver in a plurality of transmission channels, characterized in that it comprises: means for determining a distribution of probabilities of use of a transmission channel the plurality as a function of at least one value of a performance metric of the transmitter in at least one channel of the plurality; means for selection of a channel by random draw according to the distribution of probabilities of determined use. 11. Transmission device according to claim 10, characterized in that the determination means and the selection means are designed to implement a plurality of iterations comprising the determination of a probability distribution and the selection of a channel. 12. Transmission device according to claim 11, characterized in that the means for determining the probability distribution are able to determine the probability of use of a given channel to the current iteration as a function of the difference between an average value of the performance metric, measured in the given channel at each iteration between a first iteration and a second iteration prior to the current iteration, and an average value of the performance metric, measured in the selected channels between the first iteration and the second iteration. Transmitting device according to one of Claims 10 to 12, characterized in that the determination means are able to determine the distribution of utilization probabilities as a function of a plurality of values obtained by performing a measurement of at least one transmitter performance metric in each channel of the plurality of channels. 5 Transmission device according to one of Claims 10 to 13, characterized by means for determining the value of the performance metric as a function of at least one measurement performed by a receiver. 15. Computer program comprising instructions for implementing the method according to one of claims 1 to 9 when the program is executed by a processor.