Improving Channel Capacity Using Adaptive MIMO Antennas

8 downloads 486 Views 885KB Size Report
SPACE-TIME processing represents the new frontier of wireless communications [1]. Indeed, this novel tech- nology allows a better use of the spectrum resource ...
IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 54, NO. 11, NOVEMBER 2006

3481

Improving Channel Capacity Using Adaptive MIMO Antennas Marco Donald Migliore, Member, IEEE, Daniele Pinchera, Student Member, IEEE, and Fulvio Schettino, Member, IEEE

Abstract—A novel type of multiple-input multiple-output (MIMO) antenna employing parasitic elements is presented. A proper model for the parasitic-MIMO system is first discussed and then numerically and experimentally investigated. The results show that the proposed solution can significantly improve the performance of the communication system with a minimum impact on the complexity and cost of the overall system. Index Terms—Adaptive antennas, genetic algorithms, multipleinput multiple-output (MIMO) communication systems, switched parasitic antennas.

I. INTRODUCTION PACE-TIME processing represents the new frontier of wireless communications [1]. Indeed, this novel technology allows a better use of the spectrum resource, potentially making possible the gigabits per seconds rate required by the short-coming large scale diffusion of multimedia/internet mobile communication services. Two main technologies based on space-time processing are presently investigated. The first one uses adaptive antennas, a technology born at the end of the 1940’s for military applications, to improve the signal to noise and interference ratio (SINR). At the end of 90’s the Bell Labs introduced a new technology, called multiple-input multiple-output (MIMO), which permits, at least theoretically, an almost linear increasing of the channel capacity with respect to the minimum number of transmitting and receiving antennas [2]. Broadly speaking both these technologies are based on a proper handling of the signals transmitted and received by an array of antennas. Recently, another solution has been successfully investigated, based on smart antennas implementations using only one active element and a number of parasitic elements [3], [4]. Such an idea is at the base of the antenna presented in this paper: to take advantage of the parasitic adaptive antennas technology to achieve an ADAptive-Mimo antenna (AdaM antenna hereafter). In order to clarify the philosophy on which the antenna is based, it is useful to point out some characteristics of a

S

Manuscript received February 22, 2006; revised May 4, 2006. This work was supported in part by the Italian Ministry of University (MIUR) under a Program for the Development of Research of National Interest PRIN Grant 2004093025004. The authors are with the Microwave Laboratory at the DAEIMI, University of Cassino, 03043 Cassino, Italy (e-mail: [email protected]; pinchera@ unicas.it; [email protected]). Color versions of Figs. 4 and 6 are available online at http://ieeexplore.ieee. org. Digital Object Identifier 10.1109/TAP.2007.884302

standard time-domain communication system. An indubitable corner stone for the research on such systems is the work by Shannon, who introduced the concept of channel capacity [5] as a metric to characterize a communication channel. A key point in such a theory is the idea of “channel use.” According to this approach, the continuous temporal channel is equivalent to a discrete channel used a number of times each second. In case of a continuous bandlimited channel having bandwidth and signal observation time , this number turns out to be . practically equal to the time-bandwidth product On the contrary, the research on the efficient use of the “space-domain” resource is far from being a mature field. Indeed, also space, like time, is a finite and precious resource. In fact, also for space-signals, like time signals, it is possible to [6], depending on the extension define a spatial bandwidth and geometry of the sources. This allows the introduction of , wherein is related to the space-bandwidth product the spatial extension of the observation domain [7]. The spacebandwidth product is the equivalent of the time-bandwidth product for time-domain signals, and limits the “spatial channel use,” i.e., the number of times that the channel can be used in the spatial domain [7] (the space-bandwidth product is strictly related to the concept of the effective number of degrees of freedom of a MIMO channel, i.e., the number of single-input single-output (SISO) sub-channels conveying a significant amount of information [8]). The amount of information that is possible to retrieve by measuring the field depends on the number and position of the elements of the antenna. The problem of the determination of the optimal number and position of the RX antennas has been analytically solved only in some relatively simple cases, like sources and scatterers enclosed in a convex volume, and channel matrix normalized to its Frobenius norm [7]. However, in real cases regarding dynamically variable complex environments, the problem of an efficient use of the spatial resource does not allow a simple solution. A straightforward solution is to fill the available space by means of closely spaced antenna elements. However, apart from the severe mutual coupling problems, this method is cost prohibitive. A sub optimal but more affordable solution consists in the optimization of the spatial resource with a fixed hardware complexity in terms of number of receiving and/or transmitting microwave modules. A well investigated method to obtain such a goal is the so-called “element selection” (or “antenna selection”) method [2]. This strategy consists in selecting a sub-set of the available antenna elements to maximize a proper objective function, for example the information rate.

0018-926X/$20.00 © 2006 IEEE

3482

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 54, NO. 11, NOVEMBER 2006

In spite of its conceptual simplicity, the antenna selection method suffers from a number of drawbacks. The realization of an antenna based on the element selection method requires a number of well-matched and response-stable microwave switches, that increase the cost of the antenna. Furthermore, the distance between the elements must be of the order of half-wavelength to avoid strong coupling among the elements. Consequently, their dimension must be at least a couple of wavelengths. As a matter of fact, in spite of the large research on the antenna selection method (f.i. [9]–[12]) its application is limited to a small number of antennas. In this paper, a new approach for an efficient use of the spatial resource with a fixed hardware complexity in terms of receiving (or transmitting) chains is proposed. The basic idea is to introduce a number of parasitic elements around the active antennas. Each parasitic element is terminated on an electronically controllable load. The value of the loads, and consequently the antenna response, is adaptively modified in order to maximize the channel capacity. In practice in the proposed antenna the presence of passive elements makes the voltages induced at the gaps of the active elements a function not only of the field on the active antenna, but also of the field in the points wherein the passive elements are positioned. Consequently, roughly speaking the antenna is able to smartly collect information on the field not only in the position of the active elements but also in the surrounding space, so we can get advantage of a sort of “parasitic diversity.” It is interesting to note that the way of working of such antennas can also be explained from a different point of view. In fact, this antenna can be seen as a classical MIMO antenna embedded in a controllable local scattering environment. Each configuration of loads on which the parasitic elements are terminated gives a different channel matrix. The antenna finds the maximum capacity in the set of the channel matrices associated to the configurations of the loads. The use of parasitic antennas in the framework of MIMO systems was also proposed by Svantesson and Wennstrom [13], but following a different philosophy. In fact, the antenna proposed in [13] consists in a single active element rounded by a number of parasitic elements. The fast switching of the elements allows to obtain different patterns. The channel matrix is built using the signals received by these patterns. Consequently, the antenna takes advantage of pattern diversity, i.e., it can be seen as a number of antennas placed in the same spatial point, and having different patterns. On the contrary the antenna proposed in this paper uses a number of active antennas, and the passive elements are used to obtain a suitable local scatterer environment. Consequently, this approach avoids the high frequency switching that could enlarge the band occupied by the received signals. The architecture proposed in this paper is an extension of the one recently proposed for a null-forming adaptive antenna [14], and shares with it a number of advantages, like the low-cost, the simplicity and the robustness with respect to failures of the parasitic elements. Of course, other architectures, using varactors or MEMS, are possible. Indeed, all the several architectures proposed for parasitic adaptive antennas can be used as basis of AdaM antennas.

Fig. 1. AdaM system model.

Finally it has to be underlined that classical MIMO antennas need a rich scattering environment to give the best performances. On the contrary, we will show that the AdaM antenna, thanks to its “adaptivity,” is able to achieve satisfactory performances even in a poor scattering environment. The paper is organized in the following way. The numerical model of the parasitic MIMO antenna is discussed in Section II. In the following Section, a particular geometry of the AdaM antenna, consisting in two active elements and six parasitic elements, is numerically investigated considering a simple uncorrelated scattering scenario, with a number of scattering elements randomly placed around the TX and RX antennas. Section IV is then devoted to the description of the experimental results obtained indoor using the AdaM antenna numerically investigated in Sections II and III. Finally Section V summarizes the obtained results and suggests some possible developments of this novel type of antenna. II. THE NUMERICAL MODEL Let us consider a MIMO system consisting in transmitting antennas and receiving antennas (for the sake of simplicity all of them are assumed wire antennas). A scheme of principle of the system is sketched in Fig. 1, where the transmitting and the receiving arrays have been modeled with the impedance maand respectively [15]. The sources can be reptrices resented by means of Thevenin’s theorem by an ideal voltage in series with a lumped impedance . The gap source can be easily found as currents (1) where

, , . For the sake of simplicity let us imagine that the distribution of the currents over the wires depends only by the gap current; relating the in this hypothesis we can identify an operator to the vector , gap current vector i.e., the vector of the open circuit voltages on the gaps of the receiving antennas (2) It has to be stressed that the operator includes any electromagnetic effect (propagation, scattering, diffraction).

MIGLIORE et al.: IMPROVING CHANNEL CAPACITY USING ADAPTIVE MIMO

On the receiver side it is quite simple to relate the open cirto the voltages on the th receiver cuit voltage vector , obtaining impedance (3) where , and is the impedance matrix of the receiving array. and can therefore be A direct relation between readily obtained as (4) The first advantage of this formulation is that we can directly evaluate the effect of a change of the impedances values of the generators or of the receivers. Furthermore it is very simple now the modelization of the parasitic elements. In fact, if a parasitic is element is connected to the th receiving antenna, then the variable impedance controlling the element. In such a case, , so we we do not have direct access to some of the voltages can write (5) where is a permutation matrix [16] selecting the active receiving antennas, and is the voltages of the vector of the voltages at the receiver. Following the same line of reasoning, if a parasitic element is connected to the th transis the variable impedance controlling mitting antenna, then . Finally, we can write the element, and (6) where is a permutation matrix imposing a zero voltage generator on the transmitting parasitic antennas, is the number of active transmitting antennas and is the vector of the voltages at the transmitter. So we have obtained the channel matrix of the parasitic-MIMO system. Thus, we [5] of are able to evaluate the Shannon’s channel capacity our AdaM system as [2] (7)

where is the total transmitted power, is the th singular is value [16] of the matrix , is the rank of the matrix , the variance of the additive white Gaussian noise at the receiver, is the percentage of the transmitted power on the th and channel. It is clear now that the capacity is a function of the particular impedances of the switches and this introduces the parasitic diversity that can be used to improve the overall performances of the system. In order to improve the simulation of the AdaM antenna the aforementioned scheme has been slightly modified, introducing

3483

a complete MoM [17] model of the antenna elements. For the simulation of the scattering environment, since our work is not focused on the modelization of the propagation channel, we considered instead a simple uncorrelated scattering model [2]. III. CONTROL ALGORITHM AND NUMERICAL EXAMPLES A. An Evolutionary Approach The maximization of the channel capacity requires a proper algorithm to identify the configuration of the suitable loads to be connected to the parasitic antennas. The control algorithm strictly depends on the choice of the loads. In fact, two different choices can be adopted for the loads connected to the parasitic elements. A first possibility is to vary continuously the loads in a given range, for example using varactor diodes. In such a case a maximization algorithm for continuous functions is required, e.g., the conjugate gradient algorithm. A different choice is to assign a discrete number of values to the loads, using PIN diodes or MEMS to select the value of the load. In this case the maximization algorithm must operate on variables assuming only a discrete number of values. The experimental results reported in Section IV regard an AdaM antenna with loads assuming only two different values. This choice has a number of advantages, including the simplicity, the robustness and the low cost. Herewith, we will therefore investigate only maximization algorithms operating on disand the number crete variables and we will indicate with of parasitic elements at the transmitter and the receiver respec. tively, with different channel According to this choice, we can have realizations, as we have N total switches on the parasitic elements. If we assign the binary digit 0 to a switch in the OFF state and 1 to a switch in the ON state we obtain a binary vector characterizing the state of the system: it is obvious we are facing an optimization problem over a N-dimensional binary space. Since the relationship between impedances and fields is non linear and we only need to switch between two values, evolutionary algorithms are the natural candidates for maximization [18], [19]. As a first choice, the variable dynamic binary genetic algorithm (VD-BGA) described in [14], has been used as a global optimizer for our AdaM system; in the following we will refer to this as standard algorithm. On the other hand, in case of AdaM antennas used both in transmission and reception this solving method could be difficult to implement since we need to access and control all the switches simultaneously. Consequently, a slight variation of the control algorithm has been introduced: for a fixed configuration of the switches of the transmitting antenna, one step of the genetic algorithm is performed on the reduced population of the switches of the sole receiving antenna. Then, keeping the best configuration of switches found on the receiving side, one step of the algorithm is performed on the transmitting antenna and so on. In the following, the modified algorithm will be referred to as alternating algorithm, and the parameters of both standard and alternating algorithms used are summarized in Tables I and II.

3484

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 54, NO. 11, NOVEMBER 2006

TABLE I DESCRIPTION OF THE STANDARD ALGORITHM

TABLE II DESCRIPTION OF THE ALTERNATING GENETIC ALGORITHM

Fig. 2. Parasitic MIMO system: dark gray-active elements; light gray-parasitic elements.

B. Numerical Examples In the following all the dimension are normalized with respect to the working wavelength . Many different geometries have been tried for the AdaM antenna but only the numerical results relative to the geometries depicted in Fig. 2 are presented: two active elements surrounded by six parasitic elements both on the transmitting and the receiving part; every wire is about half wavelength long and the . We will consider entire structure is contained in a box of also switches with the following impedances: and . These values were chosen in

accordance with the mean values of the switches used in the experimental prototype discussed in Section IV. The proposed AdaM antenna system has been benchmarked with the sole active elements at the same position of the active elements of the AdaM antennas, and with the antenna selection approach. For the antenna selection we adopted the following rectalgorithm: we considered two (for RX and TX) angular lattices of non-coupling elements, separated from and we chose on those grids, by means of each other by exhaustive search, the couple of elements giving the maximum capacity. So we had 45 possible positions for the two TX antennas and 45 possible positions for the two RX antennas. We first consider a scattering environment of 20 point scatby terers, randomly placed in a virtual rectangular room of , with a random positioning of the receiving and transmitting antenna too; the probability density function of the distribution of the scatterers in the virtual room is uniform, but we imposed that no scatterers could be around the TX/RX antennas in a radius of . For each channel realization we first evaluated the channel in the case of sole active antennas in the open loop case, i.e., with , and we fixed the quantity to achieve a channel capacity of 4 bit/s/Hz. Then we added the parasitic elements around the active antennas, and we evaluated the channel capacity considering the . For the sake of comparison, we evaluated same value of also the channel capacity using the antenna selection method, as . described above, operating with the same In order to obtain a reasonable statistic of the performance improvement, we repeated the above steps considering 1000 different scattering scenarios.

MIGLIORE et al.: IMPROVING CHANNEL CAPACITY USING ADAPTIVE MIMO

Fig. 3. Performance comparison obtained averaging over 1000 channel realizations (equal power on each subchannel).

Fig. 4. Distribution of the capacities over 1000 channel realizations: (a) antenna selection approach-exhaustive search; (b) optimization on the RX side only-exhaustive search; (c) AdaM TX/RX system after 10 iterations-standard algorithm; (d) AdaM TX/RX system after 10 iterations-alternating algorithm.

In Fig. 3 equal power on each SISO channel has been consid, using an AdaM antenna on either TX ered and RX side, or only on the RX (or TX) side. Since only configurations of the switches are possible on the sole RX side, for this configuration an exhaustive search has been performed. The mean performances over the various environments within 10 iterations of the algorithm are shown. The use of parasitic elements on the sole RX improves the average channel capacity from 4 bit/s/Hz to more than 5.5 bit/s/Hz (dotted line in Fig. 3), while the use of AdaM antenna on both TX and RX side gives a channel capacity of almost 7 bit/s/Hz. This capacity is comparable to the one achievable using the antenna selection approach (dashed line in Fig. 3). The convergence of the algorithm is fast, and the solution is typically very good after only five iterations. Finally, it is worth

3485

Fig. 5. Performance comparison obtained averaging over 1000 channel realizations (water filling).

Fig. 6. Distribution of the capacities over 1000 channel realizations: (a) antenna selection approach-exhaustive search; (b) optimization on the RX side only-exhaustive search; (c) AdaM TX/RX system after 10 iterations-standard algorithm; (d) AdaM TX/RX system after 10 iterations-alternating algorithm.

noting that the alternating and the standard algorithm have similar performances.1 For the sake of completeness, in Fig. 4 the distribution of the capacities over the 1000 channel realizations after 10 iterations is shown. The same comparison has been performed introducing the water filling [2], [8] with which we evaluate the channel capacity of each channel realization.2 To simplify the comparison with the open loop case we considered the same 1000 channel of the previous case: the rerealization and the value of sults are shown in Figs. 5 and 6. The performances obtained by means of the introduction of the waterfilling algorithm confirm the advantages of the AdaM antenna. 1It has to be underlined that the two different algorithms are perfectly comparable since they make the same number of oracle calls. 2In the case of the genetic algorithms the waterfilling was performed in each step of the genetic algorithms’ iterations.

3486

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 54, NO. 11, NOVEMBER 2006

Fig. 7. Comparison of the mean performances of the AdaM system with the antenna selection method on both sides varying the number of scatterers (averaged on 1000 channel realizations).

Fig. 8. Comparison of the mean performances of the AdaM system with the antenna selection method on both sides varying the angular spreading of the 20 scatterers around the RX antenna (averaged on 1000 channel realizations).

It is interesting to compare the results shown in Figs. 3 and 5. Using the waterfilling algorithm, the channel capacity of the sole active antennas increases from 4 to almost 4.5, i.e., an improvement of almost 12%. However, the channel capacity in case of AdaM antenna and antenna selection are almost unchanged. This is due to the fact that, in order to take full advantage of the spatial resource, both the AdaM and the antenna selection give solutions having smaller spreading of the singular values with respect to the sole active antennas. As a consequence the equidistribution of the transmitted power on the two spatial channels is close to the optimal distribution given by the waterfilling algorithm, provided that the power level is large enough with respect to the noise level (as in the case of the examples discussed). In this condition, the switch configuration maximizing the channel capacity in case of power equidistribution is generally the same that maximizes the channel capacity in case of waterfilling. It is understood that if the transmitted power is too low, the waterfilling algorithm transmits all the available power only on the spatial channel associated to the highest singular value. In this case, we have a significant difference between the channel capacity in case of power equidistribution and waterfilling algorithm. Obviously, the AdaM switch configurations in the two cases are different. All the simulations were done considering an environment of 20 uncorrelated scatterers; such a number was taken to have rich scattering. In Fig. 7 we show a comparison between the mean performances of the AdaM system after 10 iterations and the antenna selection method in the case of water filling varying the number of scatterers: it is possible to see that for poor scattering environment (up to 10 scatterers) the AdaM system outperforms, thanks to its adaptivity, even the powerful antenna selection technique. With more than 10 scatterers the mean performances of the AdaM antenna are comparable with the ones of the antenna selection technique, but the channel capacities obtained with these two systems are almost constant with the increasing of the number of scatterers, confirming the validity of the choice of 20 scatterers.

To confirm the capability of the AdaM system to achieve good performances even in the case of non-rich scattering environment we compared the performances of the antenna selection technique and the AdaM system in the case of a variable angular spread of the scatterers around the RX antenna. We fixed and we imthe distance between TX and RX antennas to posed that for each channel realization the position of each of the 20 scatterers, relative to the RX antenna, could be modeled as a truncated laplacian distribution for the azimuthal angle and as a uniform distribution between and for the radial distance from the RX antenna. The p.d.f. of the truncated Laplacian distribution is

(8)

where B defines the variance of the distribution and A is a proper normalization constant depending on B. The angular spreading is defined as the square root of the variance of the distribution. We also imposed that for each channel realization the angular orientation of both TX and RX antennas was randomly chosen, and the value of was chosen, at each channel realization, to achieve the channel capacity of 4 bit/s/Hz in the case of the sole active antennas. The result of this comparison, shown in Fig. 8, confirms the good performance of the AdaM antenna. A large number of simulations have been performed, varying the geometry and position of the AdaM antennas, and the number and position of the scatterers, but they cannot be presented here for the lack of space. All the obtained results confirmed the validity of the proposed solution. Generally speaking the use of an AdaM antenna even on a sole side of the communication system gives a sensible improvement of the channel capacity of the link, but the best performances are achieved when AdaM antennas are used on both sides.

MIGLIORE et al.: IMPROVING CHANNEL CAPACITY USING ADAPTIVE MIMO

3487

Fig. 10. The measurement environment: TX: transmitting antenna position, RX1: receiving antenna-first position, RX2: receiving antenna-second position.

Fig. 9. The constructed AdaM antenna.

IV. EXPERIMENTAL VALIDATION A. The Prototype of AdaM Antenna We built 2 monopolar arrays (both the parasitic and the sole active one) working at 1.063 Ghz ( almost equal to 28 cm), on an aluminum ground plane of 1.24 m diameter. Every antenna long and 1.5 mm thick and is terminated is a wire about on a SMA female connector; the geometry chosen is the one in Fig. 3. The result of the construction is shown in Fig. 9. Every antenna array was placed at a height of about one meter from the floor. The passive antennas were closed on an electronic switch based on BAP 51-31 PIN diode of Phillips Semiconductors. Each switch was accurately characterized in both ON and OFF state by means of a vector network analyzer Anritsu 37217C with SOLT calibration using the Anritsu 3650 calibration kit. The mean measured impedance value of the switches was and . The switches were designed to be controlled by means of a voltage ranging from 0 to 5 V, with a very small current absorption. Consequently, the switches could be driven by a standard DAQ-24 Digital I/O PCMCIA card of the National Instruments. More details on the system can be found in [14]. B. Measurement Method and Results We measured separately each element of the channel matrix . To measure the element of the channel matrix we connected the output port of the th element of the transmitting antenna and the th element of the receiving array to the vector network analyzer Anritsu 37217C. During the measurement the other active elements were terminated on 50 dummy loads. GORE Faseflex cables were used to connect the antennas to the VNA. All the electronic parts of the set-up (electronic switches, and VNA) were controlled by means of a MatLab program running on a Laptop. The measurement of the sole active array is almost straightforward. Regarding the measurement of the AdaM system, we have different channel realizations depending on the status

Fig. 11. Measured mean performances of the AdaM SYTEM with equal power on each subchannel-first scenario.

of the switches. We measured for each configuration of the ), and switches (i.e., 4096 configurations in the case of then we performed the calculations and the optimization in a sort of off-line mode. Since no person or moving object was in the test area, no particular effort was spent to decrease the measurement time, and the 4096 channel measurements were performed in almost two hours. The first measurements have been made in the semi-anechoic chamber of the University of Cassino, in order to tune the measurement scheme in a highly controllable environment. Then we started measuring the performances of the AdaM system in a real-life environment, i.e., into a common office room (Fig. 10). We made several measurements in different positions of the antennas; in particular we measured the channel in the case of the sole active antennas and we chose the transmitted power to achieve a channel capacity of 4 bit/s/Hz, and then we evaluated the performances of the AdaM system with the same transmitted power. In Fig. 11 it is possible to see the results obtained by the two presented algorithms (averaged on 1000 sessions) with the RX antenna in the first position: both the algorithms give an average capacity greater than 7 bit/s/Hz after five iterations, and

3488

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 54, NO. 11, NOVEMBER 2006

As a last observation, it is interesting to note that the problem discussed in this paper can be reduced to an antenna synthesis problem in which the goal is represented by the maximization of the channel capacity. In some way this represents the more natural approach for antenna synthesis, since it maximizes the quantity of true interest in any communication system, that is the amount of information transmitted from the TX to the RX antenna. From this point of view antenna systems maximizing the capacity represent the natural development of smart antenna systems maximizing signal/noise and interference ratio (SINR). Information-based antenna synthesis is a new and exciting field of research. The introduction of Information Theory in the design of antennas and electromagnetic sensors appears a promising approach not only in the framework of standard communication systems, but also in other fields in which the goal is to “collect information” using electromagnetic energy, like microwave tomography and MIMO RADAR. Fig. 12. Measured mean performances of the AdaM SYTEM with equal power on each subchannel—second scenario.

ACKNOWLEDGMENT they seem almost to reach the global maximum within 10 iterations. In Fig. 12 it is possible to see the results obtained in a different position of the antennas (RX antenna in the second position). The results are always averaged on 1000 sessions of the algorithms as in the previous case. Both algorithms reach the capacity of 7 bit/s/Hz within 4 iterations when the capacity obtained by means of the sole active antennas is 4 bit/s/Hz. Finally, no significant improvement was obtained using the waterfilling algorithm, according to the observations made in Section III. Besides the results shown we made other measurements in different environments and we obtained performances comparable with the ones presented; it is worth noting that we did not perform the comparison with the maximum performances obtainable with the antenna selection technique since the exhaustive search measurement would have required too much time, but since we have had a good agreement between simulations and measurements we think that it was not a necessary step for the study. V. CONCLUSION A novel type of Adaptive-MIMO (AdaM) system has been investigated. A suitable model has been presented for the characterization of the effect of the parasitic elements, and the validity of the solution proposed has been demonstrated both numerically and experimentally. The major advantage of the proposed AdaM antenna is its simplicity with respect to the good improvement of the performances in case of both rich and poor scattering environment. Such an improvement is quite high, considering that the results of the AdaM system are comparable with the ones obtained by the antenna selection technique; furthermore it is possible to achieve good performances even if the parasitic-MIMO antenna is used only on one side of communication. Numerical and experimental results show that the algorithms require a small number of iterations, typically less than 10, to reach the maximum of the channel capacity, while a satisfactory solution for the optimization problem is reached within only 5 iterations.

The authors thank Dr. M. Di Zazzo and Dr. V. Patriarca for their precious contribution in the antennas measurements. REFERENCES [1] “Gigabit wireless,” Proc. IEEE, vol. 92, no. 2, Feb. 2004, Special Issue. [2] A. Paulraj, R. Nabar, and D. Gore, Introduction to Space-Time Wireless Communications. Cambridge, U.K.: Cambridge Univ. Press, 2003. [3] K. Gyoda and T. Ohira, “Design of electronically steerable passive array radiator (ESPAR) antennas,” in Proc. Antennas and Propagation Soc. Int. Symp., Jul. 16–21, 2000, vol. 2, pp. 922–925. [4] N. L. Scott, M. O. Leonard-Taylor, and R. G. Vaughan, “Diversity gain from a single-port adaptive antenna using switched parasitic elements illustrated with a wire and monopole prototype,” IEEE Trans. Antennas Propag., vol. 47, no. 6, pp. 1066–1070, Jun. 1999. [5] C. E. Shannon, “A mathematical theory of communication,” Bell Syst. Tech. J., no. 27, pp. 379–423, 1948, 623-656. [6] O. M. Bucci and G. Franceschetti, “On the spatial bandwidth of scattered field,” IEEE Trans. Antennas Propag., vol. AP-36, pp. 1445–1455, Dec. 1987. [7] M. D. Migliore, “On the role of the number of degrees of freedom of the field in MIMO channel,” EEE Trans. Antennas Propag., vol. AP-54, no. 2, pp. 620–628, Feb. 2006. [8] G. J. Foschini and M. J. Gans, “On the limits of wireless communications in a fading environment when using multiple antennas,” Wireless Pers. Commun., vol. 6, pp. 331–335, Mar. 1998. [9] D. A. Gore and A. J. Paulraj, “MIMO antenna subset selection with space-time coding,” IEEE Trans. Signal Processing, vol. 50, no. 10, pp. 2580–2588, Oct. 2002. [10] A. Gorokhov, D. Gore, and A. Paulraj, “Receive antenna selection for MIMO flat-fading channels: Theory and algorithms,” IEEE Trans. Information Theory, vol. 49, no. 10, pp. 2687–2696, Oct. 2003. [11] M. A. Jensen and M. L. Morris, “Efficient capacity-based antenna selection for MIMO systems,” IEEE Trans. Vehic. Technol., vol. 54, no. 1, pp. 110–116, Jan. 2005. [12] A. F. Molish, M. Z. Win, and J. H. Winters, “Capacity of MIMO systems with antenna selection,” in Proc. IEEE Int. Conf. Communications, Jun. 11–14, 2001, vol. 2, pp. 570–574. [13] M. Wennstrom and T. Svantesson, “An antenna solution for MIMO channels: the switched parasitic antenna,” in Proc. IEEE 12th Int. Symp. Personal, Indoor and Mobile Radio Communications, Sep. 30 – Oct. 3 2001, vol. 1, pp. A-159–A-163. [14] M. D. Migliore, D. Pinchera, and F. Schettino, “A simple and robust adaptive parasitic antenna,” IEEE Trans. Antennas Propag., vol. 53, no. 10, pp. 3262–3272, Oct. 2005. [15] C. A. Balanis, Antenna Theory. New York: Wiley, 2005. [16] G. H. Golub and C. F. Van Loan, Matrix Computations, 3rd ed. London, U.K.: The John Hopkins Univ. Press, 1996. [17] R. F. Harrington, Field Computation by Moment Methods. New York: Macmillan Company, 1968.

MIGLIORE et al.: IMPROVING CHANNEL CAPACITY USING ADAPTIVE MIMO

[18] T. Back, U. Hammel, and H. P. Schwefel, “Evolutionary computation: Comments on the history and current state,” IEEE Trans. Evolutionary Comput., vol. 1, no. 1, pp. 3–17, Apr. 1997. [19] J. M. Johnson and Y. Rahamat-Samii, “Genetic algorithms in engineering electromagnetics,” IEEE Antennas Propag. Mag., vol. 39, no. 4, pp. 7–21, Aug. 1997.

Marco Donald Migliore (M’04) received the Laurea degree (honors) in electronic engineering and the Ph.D. degree in electronics and computer science from the University of Napoli “Federico II,” Naples, Italy, in 1990 and 1994, respectively. He was a Researcher at the University of Napoli “Federico II” until 2001. He is currently an Associate Professor at the University of Cassino, Cassino, Italy, where he teaches adaptive antennas, radio propagation in urban area and electromagnetic fields. He teaches microwaves at the University of Napoli “Federico II.” He is also a consultant of industries in the field of advanced antenna measurement systems. His main research interests are antenna measurement techniques, MIMO and adaptive antennas, and medical and industrial applications of microwaves. Dr. Migliore is a Member of the Antenna Measurements Techniques Association (AMTA), the Italian Electromagnetic Society (SIEM), the National Inter-University Consortium for Telecommunication (CNIT) and the Electromagnetics Academy. He is listed in Marquis Who’s Who in the World, Who’s Who in Science and Engineering, and in Who’s Who in Electromagnetics.

3489

Daniele Pinchera (S’05) was born in Cassino, Italy, in 1980. He received the Dr. Eng. degree (summa cum laude) in telecommunication engineering from the University of Cassino, in 2004, where he is working toward the Ph.D. degree. His current researches are in the fields of smart antenna technologies, MIMO systems and the development of efficient evolutionary and neural computation techniques.

Fulvio Schettino (M’99) was born in Naples, Italy, in 1971. He received the Laurea degree (summa cum laude) in electronic engineering in 1997, and the Ph.D. degree in electronics and computer science in 2001, both from the University Federico II, Naples. Since June 2001, he has been a Researcher at the University of Cassino, Cassino, Italy. His main research activities concern analytical and numerical techniques for antenna and circuits analysis and adaptive antennas.