Method of searching for free band for a cognitive telecommunication terminal

[0001] This invention relates to cognitive radio systems, particularly those intended for transmission on bands allocated to the DVB-T digital video broadcasting-terrestrial system.

[0002] The increasing congestion of the spectrum has made it necessary to envisage communication systems capable of coexisting with previously allocated frequency band systems called primary systems. A large amount of research is currently being made on two coexistence strategies. The first consists of using a very low signal level due to very strong spectrum spreading, which is the method used for Ultra Wide Band (UWB) systems. The second consists of cognitively using part of the spectrum that is occasionally or temporarily unoccupied, in other words the cognitive radio method. A description of cognitive radio is given in the founding thesis by J. Mitola entitled <<Cognitive radio: an integrated agent architecture for software defined radio>>, Royal Institute of Technology, Stockholm, PhD Dissertation, May 8 2000.

[0003] The principle of cognitive radio has been used particularly for the development of Wireless Regional Area Networks (WRAN). These networks are currently undergoing a standardization procedure within the IEEE 802.22 work group. More precisely, this standard currently being produced proposes to make cognitive use of vacant UHF and VHF bands to make point to point wireless transmission in a WRAN network. In particular it is envisaged that UHF bands allocated to the DVB-T (Digital Video Broadcast-Terrestrial) television system can be used for this purpose. The DVB-T standard uses an OFDM (Orthogonal Frequency Division Multiplexing) modulation to transmit compressed video/audio streams. An introduction to IEEE standard 802.22 is given in the article by C. Cordeiro et al. entitled <<IEEE 802.22: an introduction to the first wireless standard base on cognitive radios>>, published in the Journal of Communications, Vol. 1, No. 1, April 2006, pp. 38-47.

[0004] Before being able to emit in a given UHF band, a WRAN network emitter needs to determine whether or not this band is already occupied by the primary system. It has been proposed that the autocorrelation properties of an OFDM signal can be used to detect if a DVB-T signal is present in a given band, making use of the presence of the cyclic prefix. A description of such a detection method is given in the article by Soo-Young Chang entitled <<IEEE P802.22 Wireless RANs-sensing scheme for DVB-T>>, published in November 2006. Another OFDM signal detection algorithm is described in a French application not published No. 07 55394 deposited on 31.05.2007 in the name of the applicant.

[0005] After assuring that the band is not occupied by a primary system, the emitter must still check that the band in question has not already been used by a secondary system, by means of a specific detection algorithm.

[0006] In the case of a WRAN network, terminals of the cognitive radio system themselves use an OFDM modulation. The presence of a secondary system can then be detected by the above-mentioned detection method.

[0007] In order to find a free band, a cognitive terminal must test the frequency bands allocated to the primary system sequentially, usually in increasing or decreasing order. For example, in the case of a WRAN network, the UHF channels 21 to 69 must be scanned in sequence. This scanning may be difficult and considerably reduce the network access speed.

[0008] Therefore, the purpose of this invention is to propose a fast and efficient method of searching for free band for a cognitive telecommunication terminal.

[0009] This invention is defined by a method of searching for free band using a cognitive telecommunication terminal operating in a zone covered by a primary telecommunication system using a plurality of frequency bands, according to which said terminal classifies said bands as a function of their conditional probabilities of not being occupied by the primary telecommunication system, knowing a first set of bands already occupied by the primary telecommunication system and a second set of bands not occupied by this primary telecommunication system but already used by another cognitive terminal, said terminal beginning its search by the band with the highest conditional probability.

[0010] The method advantageously uses a plurality of iterations, said first and second sets being updated during each iteration using the result of detection in the band with the highest conditional probability during this iteration.

[0011] During each iteration, the terminal measures the power of a signal from the primary telecommunication system in the band with the highest conditional probability.

[0012] If said power is greater than a predetermined threshold value, said first set is increased by said band.

[0013] If said power is less than a predetermined threshold value, the terminal may test if another cognitive terminal already uses said band.

[0014] If said band is already used by another cognitive terminal to emit, the second set is increased by said band.

[0015] If the band is not used by another cognitive terminal to emit, said band is then selected by said cognitive terminal.

[0016] Preferably, said zone is partitioned into elementary zones and, for each given band, said conditional probability of not being occupied by the primary telecommunication system, called the conditional probability of being available, is calculated from the corresponding probabilities that the different bands are available inside the different elementary zones, and the corresponding areas of these zones in which said given band is available.

[0017] In the method presented above, the primary telecommunication system may for example be a system for broadcasting DVB-T type television signals.

[0021] We will once again consider the situation of a primary telecommunication system, for example the DVB-T system and a secondary telecommunication system such as an 802.22 network.

[0022] In the remaining description, we will consider that a set of frequency bands B={b₁,b₂, . . . ,b_N} is allocated to the primary telecommunication system. This system is deployed on a given geographic zone A divided into separate elementary zones a_k such that

[0023] $A=\bigcup _k{a}_k,$
otherwise said elementary zones define a partition of the zone A. The emitters in the primary system are distributed in zone A and use frequency bands b_n, n=1, . . . ,N according to a predetermined frequency allocation plan.

[0024] As a non-limitative example, the primary telecommunication system may be composed of Terrestrial Digital Television (DTV) emitters, zone A will cover France and the elementary zones are squares with 50 km long sides.

[0025] A cognitive terminal in the secondary telecommunication system located in the zone concerned must make a search for free band in the set B to initiate or to continue a communication. In the following description, an available band means a band allocated to the primary system in which the terminal has not detected a signal from the primary system. Free band means an available band in which the cognitive terminal has not detected a signal from another terminal of the secondary system.

[0026] The basic idea of the invention is to browse through the set B, beginning with bands with the highest probability of not being used by the primary system, in other words bands with the highest probability of being available.

[0027] A detection threshold could be fixed to help detect if there is a primary system signal present in the band. For example, if the primary system is the DTV system, a threshold of −116 dBm will be set: the band will be said to be available if there is no DVB-T signal present with a power level greater than this threshold.

[0028] During its search, the cognitive terminal will probe the bands in sequence and thus obtain information about their corresponding occupancies by the primary system and if applicable, by terminals in the secondary system. Knowing the frequency allocation plan of emitters and their corresponding positions, said information provides information in terms of probability about the situation of the terminal in a specific elementary zone of A. Probabilities of the terminal being present for these different elementary zones in turn make it possible to refine the probabilities about whether or not these bands are available. Thus, from one iteration to the next, the probabilities of occupancy by the primary system are updated during each iteration, the band with the lowest probability of occupancy being selected to be tested by the terminal.

[0029] We will use the following conventions in the remaining description:

[0030] B_occ is the sub-set of B in which the terminal has detected the presence of a signal from the primary system. More precisely, if b_nεB_occ, the terminal is in a zone in which cognitive use of the b_n band would disturb reception of the primary system.

[0031] B_opp is the sub-set of B in which the terminal detects lack of a signal from the primary system, but the presence of a signal from the secondary system. More precisely, if b_nεB_opp, the terminal must be in a zone in which a cognitive use of the band b_n does not disturb reception of the primary system. However, the band in question can be used cognitively by the terminal due to the fact that it is already used by another cognitive terminal in the secondary system.

[0032] p(b_n|B_occ,B_opp) is the probability that the band b_n is available, knowing sets of bands occupied B_occ,B_opp defined above.

[0033] since the elementary zones denoted a_k form a partition of A, the conditional probability p(b_n|B_occ,B_opp) can be broken down as follows:

[0034] $\begin{array}{cc}\begin{array}{c}p\left(b_n|B_{\mathrm{occ}},B_{\mathrm{opp}}\right)=\sum _{a_k\in A}p\left(b_n|a_k,B_{\mathrm{occ}},B_{\mathrm{opp}}\right)p\left(a_k|B_{\mathrm{occ}},B_{\mathrm{opp}}\right)\\ =\sum _{a_k\in A}p\left(b_n|a_k\right)p\left(a_k|B_{\mathrm{occ}},B_{\mathrm{opp}}\right)\end{array}& \left(1\right)\end{array}$
given that p(b_n|a_k,B_occ,B_opp)=p(b_n|a_k). Once zone a_k is known, knowledge of B_occ,B_opp does not provide any additional information that can help identify available bands in this zone. The second line of (1) means that the probability that a band is available depends on probabilities of the terminal being present in the different zones a_k considering detection results B_occ,B_opp, and the probability that said band is available, for each of these zones.

[0035] More precisely, the probability p(b_n|a_k) is the probability that the band b_n is available if the terminal is in the elementary zone a_k. This probability, denoted β_n,k, is equal to:

[0036] $\begin{array}{cc}{\beta }_{n,k}=\frac{{\sigma }_k^n}{{\sigma }_k}& \left(2\right)\end{array}$
where σ_k is the area of the elementary zone a_k and σ_kⁿ is the area of a_k in which the band b_n is available.

[0037] In practice, the following procedure is used:

[0038] For each band b_iεB, knowing the positions of the emitters in the primary system emitting in the band b_i and their corresponding emission powers, the zones Ω_i in which this band b_i cannot be used cognitively are determined. More precisely, for each emitter ε using this band, the zone Ω_i(ε) is defined as the set of points in which the electrical field is greater than a given reception threshold T₁. For digital terrestrial television, a critical threshold T₁=25.4 dB(μV/m) can be used, namely T₁=−116 dBm. For example, we can determine the value of the electrical field at a DVB-T signal point by using the propagation model described by E. Sofer et al. in document IEEE 802.22 entitled <<WRAN Channel modeling>>, r6, dated 30.08.2005.

[0039] For a given elementary zone a_k and a given band b_n, a search is made for points belonging to a_k for which the band b_n is available. More precisely, a_k is discretised using small squares and the grid of points formed by the centres of these squares is considered. A test is made for each point G in this grid to see if it is in one of these zones Ω_n. If so, the band b_n is not available in the elementary square with centre G.

[0040] FIG. 1 shows an elementary zone a_k and zones Ω_n relative to two emitters ε₁ and ε₂, in this case assumed to occupy band b_n in the zone a_k. Squares in which the band b_n is not available are shown by a cross hatched pattern.

[0041] The area σ_kⁿ is nothing more than the sum of the areas of the squares of the elementary zone a_k for which band b_n is available, in other words the squares that are left blank.

[0042] It is important to note that the calculation of conditional probabilities p(b_n|a_k) can be made once for all once the positions and the corresponding powers of the primary system emitters are known.

[0043] The second conditional probability that arises in expression (1) is the probability that the terminal is in zone a_k, knowing that:

• • [0044] (a) The terminal has detected a signal from the primary system in the b_nεB_occ bands and
  • [0045] (b) It has checked that there is no signal from the primary system in the b_nεB_opp bands and detected the presence of signal from the secondary system.

[0046] This probability can be expressed using Bayes theorem:

[0047] $\begin{array}{cc}p\left(a_k|B_{\mathrm{occ}},B_{\mathrm{opp}}\right)=\frac{p\left(a_k\right)p\left(B_{\mathrm{occ}},B_{\mathrm{opp}}|a_k\right)}{p\left(B_{\mathrm{occ}},B_{\mathrm{opp}}\right)}& \left(3\right)\end{array}$

[0048] Assuming that events {b_iεB_opp|a_k} and {b_jεB_opp|a_k} are independent if i≠j, the expression p(B_occ,B_opp|a_k) can be factorized in the form:

[0049] $\begin{array}{cc}p\left(B_{\mathrm{occ}},B_{\mathrm{opp}}|a_k\right)=\prod _{b_i\in B_{\mathrm{occ}}}p\left(b_i|a_k\right)\prod _{b_j\in B_{\mathrm{opp}}}p\left(b_j|a_k\right)& \left(4\right)\end{array}$

[0050] The probability p(b_i|a_k) for b_iεB_occ is the probability that the band b_i is used by the primary system knowing that the terminal is in the zone a_k, in other words the probability that the terminal cannot use band b_i because it is occupied by the primary system, namely p(b_i|a_k)=1−β_i,k.

[0051] similarly, the probability p(b_j|a_k) for b_jεB_opp is the probability that the band b_j is available in zone a_k. Therefore it is equal to the probability β_j,k.

[0052] The result is that the conditional probability p(a_k|B_occ,B_opp) can be expressed as follows:

[0053] $\begin{array}{cc}p\left(a_k|B_{\mathrm{occ}},B_{\mathrm{opp}}\right)=\frac{1}{p\left(B_{\mathrm{occ}},B_{\mathrm{opp}}\right)}\frac{{\sigma }_k}{{\sigma }_A}\prod _{b_i\in B_{\mathrm{occ}}}\left(1-{\beta }_{i,k}\right)\prod _{b_j\in B_{\mathrm{opp}}}{\beta }_{j,k}& \left(5\right)\end{array}$
where

[0054] ${\sigma }_A=\sum _{i=1}^K{\sigma }_i$
is the total area, and K is the number of elementary zones.

[0055] The area σ_A and the probability p(B_occ,B_opp) in (5) are common factors that do not depend on a_k. Finally, if we normalise the probability p(a_k|B_occ,B_opp) by the probability that the terminal is located in an arbitrary zone, knowing B_occ,B_opp:

[0056] $\begin{array}{cc}p\left(a_k|B_{\mathrm{occ}},B_{\mathrm{opp}}\right)=\frac{{\sigma }_k\prod _{b_i\in B_{\mathrm{occ}}}\left(1-{\beta }_{i,k}\right)\prod _{b_j\in B_{\mathrm{opp}}}{\beta }_{j,k}}{\sum _{l}p\left(a_l|B_{\mathrm{occ}},B_{\mathrm{opp}}\right)}& \left(6\right)\end{array}$
and therefore, by substituting expressions (2) and (6) in (1):

[0057] $\begin{array}{cc}p\left(b_n|B_{\mathrm{occ}},B_{\mathrm{opp}}\right)=\sum _k\frac{{\sigma }_k^n\prod _{b_i\in B_{\mathrm{occ}}}\left(1-{\beta }_{i,k}\right)\prod _{b_j\in B_{\mathrm{opp}}}{\beta }_{j,k}}{\sum _{l}p\left(a_l|B_{\mathrm{occ}},B_{\mathrm{opp}}\right)}& \left(7\right)\end{array}$

[0058] The terms σ_k^j, β_i,k for i=1, . . . ,N and k=1, . . . ,K can be determined once for all because they are only dependent on deployment of the primary network in zone A.

[0059] A cognitive terminal located in zone A uses conditional probabilities p(b_n|B_occ,B_opp) of the different bands of B to search for free band as shown in FIG. 2.

[0060] The search method begins with an initialization step, 210:

[0061] B={b₁,b₂, . . . ,b_N} is the set of bands that the primary system can use;

[0062] B_occ and B_opp are initialized as empty sets.

[0063] In step 220, it is tested if Card(B)>0. If so, we go on to calculation step 230. If not, step 225 terminates with an observation of failure: the cognitive terminal cannot use any of the bands in B to transmit.

[0064] In step 230, the terminal calculates the probabilities p(b_n|B_occ,B_opp) starting from current sets B_occ,B_opp and values σ_kⁱ, β_i,k for i=1, . . . ,N and k=1, . . . ,K stored in a memory 235.

[0065] In step 240, the probabilities p(b_n|B_occ,B_opp) are sorted and the terminal determines:

[0066] $\begin{array}{cc}{b}_{\max }=\underset{n}{\text{arg}\max }\left(p\left(b_n|B_{\mathrm{occ}},B_{\mathrm{opp}}\right)\right)& \left(8\right)\end{array}$

[0067] In step 250, the terminal detects if the band b_max is occupied by the primary system. More precisely:

• • [0068] if p(b_max|B_occ,B_opp)=1, the terminal concludes that the band b_max is not occupied by the primary system. This corresponds to a situation in which coverage zones of emitters using b_max are sufficiently far from the location at which the terminal is located;
  • [0069] otherwise, the terminal detects if the primary signal level received in this band is less than a predetermined threshold value (for example if the primary system is the DTV system and if the emitted signals are DVB-T signals, a threshold value of −116 dBm can be used). If not, it will be assumed that the band b_max is occupied by the primary system and we go onto the update step 255. On the other hand, if the band b_max is not occupied by the primary system, in other words if b_max is available, we go on to detection step 260.

[0070] In step 260, we test if the band b_max is already used by another cognitive terminal. If not, the terminal has found a free band and the algorithm stops at 270. Conversely, we go on to update step 265.

[0071] In step 255, B is updated by B=B−{b_max}, and B_occ is updated, by setting B_occ={B_occ∪b_max};

[0072] In step 265, the set B is updated by B=B−{b_max} and B_opp is updated, by setting B_opp=B_opp∪{b_max}.

[0073] In both cases, one goes back to test step 220.

[0074] The method to search for free band may be repeated periodically, particularly if the cognitive terminal is mobile.