"A SYSTEM ADAPTED FOR COMMUNICATION IN A TWO-HOP WIRELESS COMMUNICATION NETWORK"

corresponds to the
common transmission parameter, and to the
relative transmission parameter for relay station k.
The maximum attainable receiver SNR under aggregate relay transmit power constraint can be determined to
At closer inspection, it is noted that the SNR contribution from each individual relay to Har lay station would transmit with all relay transmit power jPfts themselves.
Moreover, "derivation of analytic expressions", expressions for a combination of regenerative and non-regenerative coherent combining is also presented. When studying regenerative and non-regenerative coherent combining an interesting observation is that a regenerative approach is generally inferior to non-regenerative case, because regenerative relaying by necessity is constrained to a region around the transmitter and cannot exploit all available relays in an optimal manner. With other words, even though a signal may not be decoded, it may still contribute when coherent combining is employed. In any case, a combination of non-regenerative and regenerative scheme will perform slightly better than if only the non-regenerative method is considered. The mechanisms for power and phase control that are discussed in the following are is independent and generic to whether regenerative relaying is employed as well.
Phase control
As the first implementation example the logical architecture and the method according to the present invention is adapted for the use of facilitating coherent combining. A prerequisite for coherent combining is that signals are phase-aligned at the receiver. This is enabled by compensating for the complex phase from the transmitter 210 to the relay station 215 as well as the complex phase form the relay station 215 to the receiver 220. Practically, in each relay
station the received signal, yk, is multiplied with the phase factor where argfe} = -!
Therefore, explicit or implicit channel phase information must be made available at each individual relay station. There are essential two basic schemes that can be used in deriving phase information, one based on closed loop control and one on open loop control. The closed loop control is necessary to use when channel reciprocity cannot be exploited, such as in FDD (used over a single link), or when high control accuracy is required. The open loop control scheme instead exploits channel reciprocity, e.g. enabled by TDD (used over a single link)

with channel sounding that operates within channel coherence time. Open loop control is generally less accurate than closed loop control, due to asymmetries in the transmit/receive chains for a station. The differences boils down to the effort put into hardware design, and can always be compensated by improved design. Also, incorporating occasional closed loop control cycles may compensate for static open loop errors. However, in the present invention the phase error can in principle be up to ±90 degrees and still combine coherently (but not very efficiently) with other relayed signals. Hence, absolute phase accuracy is not a must, but certainly preferred. A closed control scheme generally relies on explicit signalling, reporting the result of measurements and therefore consumes more communication resources and incurs latency relative an open loop scheme. Note that this discussion on TDD vs. FDD considers duplexing technique over a single link at a tune, e.g. the relay station to receiver link, whereas it is also possible to characterize the overall communication in the network on basis of time and frequency division. For example, link one and link two may share a frequency band or use different ones. From point of view of the invention, however, any combination of duplexing and multiple access schemes may be used, as long as channel phase information can be determined and used for phase compensation in the relay stations.
Tightly connected with closed loop and open loop control is the issue which station sends the pilots, which has been discussed previously in reference to table L Since it is the relay stations that must perform phase adjustment, this is the natural place to determine arg{ak}. If
a relay station sends a pilot signal, the phase (or channel) parameters need to be reported back to the relay. This corresponds to the closed loop case. If a relay station instead receives a pilot, the phase (or channel) parameter does not need to be reported anywhere. This corresponds to the open loop case. It is clear that depending whether phase (i.e. channel) information need to be sent away in a control packet or can be kept hi the same station, this has an impact on radio resource efficiency, power consumption as well implementation complexity. In any case, as seen form above, a myriad of possibilities exist and we select the most promising. A preferred combination of duplexing and multiple access will be further discussed. However, as appreciated by the skilled in the art a very large number of possibilities exist and the invention is not limited to the below exemplified.
Case one (see table 1), which is of open loop type and suitable for TDD with "sufficient" coherence time, offers the lowest signalling complexity as only two transmissions are necessary and the processing is distributed on all relay stations. Here, the transmitter as well as the intended receiver issue channel estimation symbols often enough or whenever needed

such that each relay can track both channels. The relay station subsequently estimates the channel phases that determine the phase factor of ak.
Power control
A second important aspect for resource efficient communication, apart from phase control, is power control, since it provides means to ensure satisfactory communication quality. The logical architecture and the method according to the present invention is readily adapted to be used for an effective power control. The power control method is based on that the effective SNR at the receiver is controlled towards a target SNR, F0, which assert the desired link
quality. The target SNR may of course change with time depending on how link mode or QoS requirement changes with time. According to the logical architecture and the method according to the present invention power may be adjusted at the transmitter and individually at each relay. The relay power control has common as well as individual relay component. In the objective of minimizing the aggregate power addresses the issue of multiple access interference minimizations as well as minimizing relay power consumption. However, when a MS act as a transmitter, the power control may also be use as a method for significantly minimizing power consumption and radiated power for the MS, which among other advantages prolongs the battery life of the MS.
On the highest level, the power control problem may be defined as: Find [PBS , P } , Vfr e {l,2,..., K} ; such that iffi = T0
This is preferably accomplished under some constraints, such as minimization of P&y = £ Pk and with fixed PBS, but other constraints may also be considered, e.g. minimization of the total transmit power P + P or by taking localization of relay induced interference generation into account. In the following, we assume minimization of P^ = £ Pk with fixed (or relatively slow) adaptation of Pss. This is a reasonable design objective in downlink, but
for uplink it may be of greater interest to minimize the transmitter power. However, if the relays are mobile and relay on battery power, the sum power of relays and transmitter may be minimized.
This is the basic function of power control. From practical viewpoint, the overall task of controlling power hi a cooperative relay network in general, and with coherent combining in

particular, is to use previous knowledge of used power PBS and Pk and update those parameters to meet desired communication quality.
Power control share much of its traits with the phase control as the gain of the links may be estimated in several ways, depending on closed/open loop, TDD/FDD, distribution of control aspects. Hence, also here can a range of alternative implementations be envisioned, hi the following, similar to the phase control discussion, it is assumed that the transmitter and receiver issue channel estimation signals and that channel gain reciprocity can be assumed, but the invention is not limited hereto.
The power control being proposed here has both a distributed component for each relay station, the relative transmission parameter, and a component common to all relays, the common transmission parameter. The scheme operates as follows: Through channel estimation, and with knowledge of the power used to send the pilot, each relay station may determine its respective path gain towards the transmitter and receiver respectively, but also interference and noise levels may be estimated at the same time. Based on path gain
measurement, and information about P^and cr^s, it is possible to determine TMS,k. Possibly also based on path gain, noise with interference estimations and PBS awareness, or simply direct SNR measurements on any received signal, the SNR at the relay station, F
can be determined. Based on this, the relative transmit power levels can be determined at each relay station in a fully distributed manner. However, each relative transmit power level need to be scaled with normalization factor pto ensure that aggregate transmit power is identical,
or at least close, to the aggregate transmit power P. This is the common power control part. If (p is too small, then more power than optimum P is sent, and hence a more optimal relative power allocation exist for the invested transmit power. The same is valid when p is too large. Hence, it is important for optimal resource investment to control p such that the intended power P is the aggregate transmit power level by the relays. N.B., it is not a significant problem from performance point of view if q is somewhat to small as that only improves the effective SNR, since the relative impact of receiver internal noise is reduced.
Referring now to the logical architecture illustrated in FIG. 5 the normalization factor, being a common transmission parameter, is preferably determined, as well as distributed from, the receiver. This should be seen as a logical architecture, since it is also possible to forward all control information to the transmitter, which then redistribute it to the relay stations, fore

example. The first control loop 505 between the receiver 220 and the relay stations 215 :k, provides the relay stations with the P, whereas the second control loop 510 from the
receiver 220 to the transmitter 210, provides the transmitter with PBS . Optionally, if the
transmitter has a better view of the whole radio system including many groups of cooperative TX-RS-RX links, similar to what a backbone connected basestation in a cellular system would have, then it may incorporate additional aspects that strive to optimize the whole system.
One method to implement the control loop at the receiver is now given, then assuming that PBS is fixed (or controlled slowly). From a transmission, occurring at tune denoted by n , the receiver measure the power of the coherently combined signal of interest, Cr , the relay induced noise measured at the receiver, Nr , and the internal noise in the receiver Nt . Based
on this, and conditioned r0 , the receiver determines P+l and an update of a normalization factor, . This can be written as a mapping through an objective function f as
Formula Removed
The receiver then distributed the updates, P(n+1) and (n+1) to all relays through a multicast control message. To illustrate the idea, assume that P^gis kept fixed from previous
transmission, but the normalization factor is to be adapted, hi the section "Derivation of analytic expression" it is shown that optimum normalization requires a balance between received signal, Cr , and the total received noise, interference and receiver internal noise Nr + Nt according to
Hence, by including the previous normalization factor, which is known by the receiver, and the update needed(n+1) to balance the equation, the relation becomes
Formula Removed
, which yields p by solving a simple second order equation.

If both PXS and (p need to be updated, the balance equation above, the relation for the receiver SNR, F, can be used together with measured signal levels and solve for P and p. Linearization techniques, such as Taylor expansion and differentials, may preferably be used for this purpose and solving for APy and A
It is noted that for the first transmission, the normalization factor is not given a priori. Different strategies may be taken to quickly adapt the power. For instance, an upper transmit power limit may initially be determined by each relay as they can be made aware of F0 and
also can determine their (coherent combining) SNR contribution. If each relay stays well below this upper limit with some factor, power can be ramped up successively by the control loop so ongoing communications are not suddenly interfered with. This allows control loops, for other communication stations, to adapt to the new interference sources in a distributed and controlled manner.
Also note that even though transmit power limitations occur in any relays, the power control loop ensures that SNR is maximized under all conditions.
Another, possibly more precise, method to determine the normalization factor is to determine the \ak term in each relay and then send it to the receiver where £ \ak \ , is calculated and hence yielding the normalization foctor p. Subsequently (p is distributed to all relays, similar to previous embodiment. Note that the amount of signalling may be reduced and kept on an acceptable level by sampling only a subset of all relays, i.e. some of the most important
relays, hi order to produce a sufficient good estimate of £ \ motivated that the $] \a^ \ term will generally not change much over short time, even hi fading channels, due to large diversity gains inherent in the invention.
Although power control has been described in the context of coherent combining, the framework is also applicable for power control hi other types of relay cooperation schemes, such as various relay induced transmit diversity, such as Alamouti diversity. The framework is similar in that the power control considers combinations of transmitter power, individual relay power and aggregate relay power. Another example of relay induced transmit diversity is (cyclic/linear) delay diversity. Each relay imposes a random or controlled linear (or cyclic) delay on the relayed signals, and hence causes artificial frequency selectivity. Delay diversity is a well known transmit diversity from CDMA and OFDM based communication.

To summarize this section, this invention suggests using power control as a concept to ensure performance optimization for coherent combining based cooperative relaying in a realistic channel and in particular to optimize signal to noise ratio under aggregate relay transmit power constraints. This power control concept is not limited to coherent combining based cooperative relaying networks, but also other cooperative relaying oriented networks may use the same concept, though then with optimization objectives most suitable to the scheme being used. In addition, the basic features for a protocol based on channel sounding and estimation of gain parameters over both link one and link two are suggested. A reasonable design choice for protocol design (with commonalities with the phase control) has also been outlined, based on low complexity, low signalling overhead and low total power consumption. In particular, it is shown that combination of power control loops including relay and transmitter power control may be used. Lastly, it has been demonstrated that the control loop for the relays may be build on distributed power control decisions hi each relay as well as a common power control part, where the whole set of relays are jointly controlled.
The main steps of the embodiment using the inventive method and architecture for efficient power control and phase control are illustrated in the flowchart of FIG. 6. The method comprises the steps of:
600: Send pilots on the k paths of link 1, from transmitter 210' to relay stations 2l5:k; 610: Each relay station 2l5:k estimates the k channel of link 1, h ; Also interference and noise levels are estimated in order to calculate
620: Send pilots on the k paths of link 2, from receiver 220' to relay stations 2 1 5 :k,
630: Each relay station 2l5:k estimates its respective channel out of the k channel of 2, h2ik ',
640: Each relay station .2 1 5 :k determines relative transmission parameters based on the channel estimates.
650: The receiver 220' determines a normalization factor p .
660: The receiver 220' broadcast the normalization factor PRS and 2RS to the relay stations 215:k

670: Each relay station 215:k uses the broadcasted PRS, and the locally determined TMS,k andTRS,K phase of channel estimates h1,k, h2,k to, on the reception of signal yk, transmit the following signal:
Formula Removed
wherein the parameters is calculated based on the channel estimate, PBS, and cr|y , and
If the first transmission to the receiver is considered, (then the power loop is unaware of the forthcoming link quality), by way of example the relay may modify and upper limit the
received normalization factor p such that , where c 1 being sent from the receiver or is a priori known.
675: The receiver 220' feedbacks control information to the transmitter 210' (.Pas).
The first control loop, indicated in step 660 may further comprise the substeps of:
660:1 The receiver measure at time n, the quality of the received signal, or more specifically the power of the coherently combined signal, Cr , the relay induced noise measured at the receiver, Nr , and the internal noise in the receiver Ni.
660:2 The receiver determines based on the measurement of step 675:1, and conditioned a desired F0 target, an update of at least one of the normalization factor,(n+1) and the
aggregate relay power P(n+1)RS.
660:3 The receiver distributed the updates, P(n+1)RS and (n+1) to all relays through a multicast control message.
Similarly, the second control loop, indicated in step 675, may optionally comprise: 675:1 The receiver update the transmitter (BS) power Pl) .

Alternatively, if no estimations and calculations are to be done by the relay stations, unprocessed results of the pilots are forwarded to a centralized functionality, in the receiver for example, and relevant transmission parameters transmitted to each relay station.
Relay stations activation control
The method and architecture of the present invention may advantageously be used for deciding which relay stations 215:£to include in a communication, either at the establishment of the communication or during the communication session. As some relays experiencing poor SNR conditions on either link (transmitter-relay and relay-receiver) or both, they may contribute very little to the overall SNR improvements. Yet, those relays may still consume significant power due to receiver, transmitter and signal processing functions. It may also be of interest to have some control means to localize relay interference generation to fewer relays. Hence, it may therefore be considered to be wasteful to use some of the relay stations. Consequently, one desirable function is to activate relays based on predetermined criteria. Such criteria may be a preset lower threshold of acceptable SNR on either link, both links or the contribution to the effective SNR. The limit may also be adaptable and controlled by some entity, preferably the receiving station as it has information on momentary effective SNR. The relay may hence, e.g. together with power control information and cannel estimation symbols, receive a relay activation SNR threshold Tcttye from the receiver to which the expected SNR contribution is compared against, and if exceeding the threshold, transmission is allowed, else not. The relay activation SNR threshold TActive corresponds to a common transmission
parameter, preferably determined by the receiver 220' and distributed to the relay station 215. The actual decision process, in which each relay station uses local parameters (corresponding to the relative transmission parameters) is distributed to the relay stations hi the manner provided by the inventive method and architecture. This test, preferably performed in each relay prior to transmission, may e.g. be formulated according to:
Formula Removed
, but other conditions, depending on relay methods including alternative relay diversity techniques, can also be used. For instance, the relay activation condition may more generally be characterized as an objective function f2 according to
Formula Removed

Moreover, The broadcasted message containing the TActive could further comprise fields that may be used to pinpoint specific relays (through assigned relay addresses) that should be incorporated, or is only allowed to be used, or must excluded or any combination thereof. Other methods to address certain relays may e.g. be based on address ranges. This enables one to limit the number of involved relays as desired.
From the above discussion and expression (9) it can be noted mat the receiver 220' may, upon experiencing weakening SNR, for example due to the movement of the MS, choose to order a increased transmission power and/or to include more relay stations 215 by lowering the threshold f^cove • Other communication quality conditions, such as packet or bit error rate, may also be used by the receiver to trigger changes in the common parameters, such as a joint transmit power scaling of all relay powers.
Relay activation control may be incorporated in the power and phase control algorithm described with reference to FIG. 6, by modifying the steps 650-670, so that:
in 650: the receiver 220' also determines an activation SNR threshold TActive in 660: the receiver 220' also broadcast rctive to the relay stations 2l5:k.
in 670: each relay station 2l5:k firstly determines if to broadcast using the activation SNR threshold TQtive, for example according to expression (9)
The method and architecture according the present invention may be adapted to other topologies than the above exemplified. The topology in FIG. 5 may, for example, be modified to include multiple antennas in each relay station as shown in FIG. 7. The benefit in doing that is that the number of relay stations can be reduced while still getting similar total antenna directivity gain. If each antenna element is separated more than the coherence distance, diversity gain is also provided. In all, this can reduce the cost, while providing near identical performance. However, reducing the number of relays may have a detrimental impact due to shadowing (i.e. log normal fading) and must be carefully applied. From signal, processing and protocol point of view, each antenna can be treated as a separate relay station. Another benefit of this approach is however that internal and other resources and may be shared. Moreover,

relaying may potentially be internally coordinated among the antennas, thereby mitigate ; interference generation towards unintended receivers.
The communication quality may be further improved by also incorporate the direct signal from the transmitter 21 0 to the receiver 220. There are at least two conceivable main methods to incorporate the signal from the transmitter. FIG. 8, depicts the topology when direct transmission from the transmitter is also considered.
In the first method, two communication phases are required. The receiver combines the signal received directly from the transmitter, in the first phase, with the relay transmission, from the second phase. This is somewhat similar to receiver based combining in the classical relay channel, but with coherent combining based relaying. Maximum ratio or interference rejection combining may be employed.
In the second method, Transmit-relay oriented Coherent Combining, only one communication phase is used, and used for coherent combining of the direct signal from the transmitter to the receiver with the relay signals. This can be made possible if relays can transmit and receive concurrently, e.g. over separated antennas. The phase of ak must then ensure alignment of relayed signal with the direct signal as
Formula Removed
, where hBS consequence of incorporating the direct signal for coherent combining is that the relays must adaptively adjust their phase relative the direct signal. A closed loop can be used for this. Similar to the normalization factor power control, the receiver issues phase control messages to the whole group of relay stations, but with a delta phase 9 to subtract from the calculated phase compensation
Formula Removed
As the basestation does not induce any noise through its transmission, its transmit power does not need to be adjusted for optimal performance as was needed for the relays. Instead, performance increases rnonotonically with increasing basestation transmit power. One option is however to try to minimize the overall transmit power, aggregate relay power and basestation power. The parameter setting for this is similar to what has been derived in the discussion on regenerative relaying, assuming that the basestation is considered as a relay. In addition to above, multiple antenna elements at the transmitter may also be used, similar to the discussions on relays with multiple antennas.

The derivation of the relative and common transmission parameters is also directly applicable to multi carrier transmission, such as OFDM by handling each subcarrier independently. This will then include a common amplitude normalization, phase and distributed relay amplitude compensation per subcarrier. For doing this, the path over FFT-processing-IFFT is taken, or possible through time domain filtering. The power control may send a normalization factory
and relay power indication P^in vector form to optimize performance per subcarrier. A more practical solution, is to send q> and P/^as scalars, acting on all subcarriers. In case of
subcarrier optimization, the power control may then try to minimize power the total transmit power over all subcarriers to meet desired communication quality. This then provides some diversity gain in the frequency domain.
Another OFDM aspect is that it is a preferred choice for the transmit-relay oriented Coherent Combining described above. The reason is that the cyclic prefix allow for some short relay transfer latency, where phase and amplitude is modified through a time domain filter enabling immediate transmission.
For single carrier transmissions, such as CDMA, and with frequency selective channels, a frequency domain operation similar to OFDM may be employed or optionally the phase alignment can be performed on the strongest signal path, or with a time domain filter as discussed for OFDM.
For coherent combining to work, it is important to synchronize relay station frequency to a common source. In a cellular system, the BS is a natural source as since the clock accuracy is generally better at the basestation than in any mobile station. This function can exploit the regular frequency offset compensation as performed in traditional OFDM receiver implementations, that mitigates niter channel interference.
However, the relays may optionally exploit GPS for frequency synchronization, if available.
While the invention has primarily been described in a context of coherent combining, the invention is not limited hereto. The invention may be applied on various types of existing and foreseen methods for 2-hop (cooperative) relaying, hi the most general case, the transmit parameters of the relays are functions of communication characteristics of the first link, communication characteristics of the second link, or a combination thereof. The communication quality has been described outgoing from complex channel gam (suitable for coherent combining), however when other schemes are considered (offering diversity and/or spatial multiplex gains), other link characteristic metrics may be of more relevance. As an

example, for Alamouti diversity it may be more preferable to use average path gain metric ,G, instead of complex channel gains, h.
While the invention has been described in connection with what is presently considered to be the most practical and preferred embodiments, it is to be understood that the invention is not to be limited to the disclosed embodiments, but on the contrary, is intended to cover various modifications and equivalent arrangements included within the spirit and scope of the appended claims.

Detailed derivation
In the analysis we assume that there are K relay stations arbitrarily located. Each relay station
K E{1,2...,K}receives a signal composed of an attenuated version of the desired signal, e.g. modeled as complex Gaussian x ~ N(0,1)»as well as a noise plus interference term,nRS,k according to
Formula Removed
, where h1k is the complex path gain from the basestation to relay station k and PBS is the transmit power of the basestation.
In the relay, yk is (for analytical tractability) normalized to unit power, and multiplied with a complex factor that generates output zk. Subsequently zk is sent over link two, towards the receiver and is on its way attenuated with complex path gain hz2,A, where it is super-positioned with signals from other relays and noise and interference is added.
As it is assumed that each relay normalize the received power plus noise to unit power prior amplification and phase adjustment, the relay transmit power constraint can be incorporated in the analysis by letting each station k use transmit power
Formula Removed
, where PRS is the total transmit power of all relay stations, and ak is a un-normalized complex gain factor for relay station k.
For aggregate power constrained relay transmission, the SNR at the receiver (Mobile Station, MS, assumed here) may then be written
f, where crMS is the noise plus interference level at the mobile station.
A condition for coherent combining is phase alignment of signals, which can be achieved by ensuring
where c} is an arbitrary constant
The expression for the effective SNR resulting from coherent combining may then be rewritten as
Note that TMSk is a "virtual SNR" hi the sense that it is the SNR if relay station k would use all aggregate relay stations transmit power by itself. It is noticed that the SNR expression has the form
which can be transformed by using
which yields
Now, the nominator is upper limited by Cauchy-Schwarz's inequality hence for an optimal bk equality can be attained and the resulting SNR is then

This may be conveniently expressed hi SNRs as
Through identification, it is seen that the maximum SNR can be attained if
Formula Removed
, where Const is an arbitrary constant that can be set to one for convenience.
From power control perspective, it is interesting to note that the nominator is exactly the square of the denominator for optimum SNR. This knowledge can therefore be used as a power control objective.
Using the reverse transformation, one yields
Formula Removed

Hence a relay receiving a signal yk can determine zk by determining
Formula Removed

Regenerative relaying add-on
If the SNR at a relay station is high enough, the received signal may be decoded prior relaying the signal. To model this behavior, let's say that larger than a minimum SNR, FDecode , is
sufficient for decoding. The benefit in doing this, is that forwarding of detrimental noise (and interference) can be avoided all together, and hence result in a further enhanced SNR at the receiver. In this case however, the decoded signal should be phase compensated only for the second hop, i.e
Formula Removed
By setting 2RS, K = 0 for those stations in the previous expressions, one can derive the magnitude of the multiplicative factor ja^ | as well as the contribution to the SNR improvement The combination of both noise-free (regenerative) and noisy (non-regenerative)
transmission then takes the form

and
Note that r is only a model useful to assess performance in a mixed non-regenerative and regenerative relaying scenario. In practice, the upper expressions, .i.e. corresponding to FTDecode, are used when the signal is not forwarded in a non-regenerative manner, and the lower expressions, .i.e. corresponding to T rDecode, are used when the signal is not forwarded in a regenerative manner.

References
[1]. J. N. Laneman, Cooperative Diversity in Wireless Networks:
Algorithms and Architectures, Ph.D. Thesis, Massachusetts Institute of Technology, Cambridge, MA, August 2002.
[2]. J. N. Laneman and G. W. Wornell, "An Efficient Protocol for Realizing Distributed Spatial Diversity in Wireless Ad-Hoc Networks," in Proc. ofARL FedLab Symposium on Advanced Telecommunications and Information Distribution (ATIRP-2001), (College Park, MD), March 2001.
[3]. J. N. Laneman and G. W. Wornell, "Energy-Efficient Antenna-Sharing and Relaying for Wireless Networks," in Proc. IEEE Wireless Communications and Networking Conference (WCNC-2000), (Chicago, IL), September 2000.
[4]. B. Schein and R. Gallagher, "The Gaussian parallel relay network," in IEEE International Symposium on Information Theory ISIT2000, Sorrento, Italy, June 25-30,2000.
[5]. B. Schein. "Distributed Coordination in Network Information Theory." PhD thesis, pp. 64-68, MIT, Cambridge, MA, August 2001. Lippman. Bletsas
[6]. T.M. Cover and A.A. El Gamal, "Capacity theorems for the relay channel," IEEE Trans. Inform. Theory, vol. 25, no. 5, pp. 572-584, Sept. 1979.
[7]. E. V. D. Meulen, "Three-terminal communication channels," Advances in Applied Prob- ability, vol. 3, pp. 120 154,1971.
[8]. A. Sendonaris, E. Erkip, B. Aazhang, "Increasing CDMA Cell
Capacity via In-Cell User Cooperation", Department of Electrical and Computer Engineering, Rice University, Poster Titles, November 11,1997.
[9]. M. Dohler, E. Lefranc, H. Aghvami, "Virtual Antenna Arrays for Future Wireless Mobile Communication Systems," ICT2002, June 2002
[10]. G. W. Wornell, V. Poor, "Wireless Communications: Signal
Processing Perspectives (Prentice Hall Signal Processing Series) Prentice Hall; 1st edition (April 1998).







We claim:
1. A system adapted for communication in a two-hop wireless communication network, wherein the network comprises a transmitter (210), a receiver (220) and at least one relay station (215), wherein the relay station (215) is adapted to forwarding signals from a first link between the transmitter (210) and the relay station (215) to a second link between the relay stations (215) and the receiver (220) characterised in that the relay station (215) uses radio channel characteristics of both the first and second link for the forwarding on the second link.
2. The system as claimed in claim 1, wherein the relay station (215) is further provided with means for performing channel characterization (216) and means for determining relative transmission parameters (217) based on channel characterization and the forwarding is at least partly based on said relative transmission parameters.
3. The system as claimed in claim 1, wherein the system is provided with means for determining a common transmission parameter which is based on the total communication quality between the transmitter (210') and the receiver (220'), and the relay station (215) is further provided with means for receiving the common transmission parameter and the forwarding on the second link is at least partly based on said relative transmission parameters and said common transmission parameter.