2012 International Conference on Control, Automation and Information Sciences (ICCAIS)

TA03

Distributed Power Adjustment in Cognitive Radio Networks Using PID Control

Genki Matsui, Yukinori Nakamura, Takuji Tachibana, Kenji Sugimoto, and Shinji Wakui
Abstract In this paper, we propose a distributed power control algorithm based on PID control and dynamic constraint condition for cognitive radio networks. In our proposed algorithm, at rst, each transmitter in secondary networks receives the feedback information about signal-to-interference plus noise ratio (SINR) from its corresponding receiver. Based on the information, the transmitter derives the transmitted power with PID control in order to satisfy the quality of service (QoS) constraints in secondary networks. Then, the transmitter decides the actual transmitted power according to a constraint condition in addition to the derived power. This constraint condition is effective for avoiding the interference with primary networks. Since the constraint condition affects the performance of our proposed method signicantly, we propose an effective update algorithm. In the update algorithm, the transmitter changes the constraint condition by comparing the derived transmitted power with both the maximum interference power tolerance and the target SINR. We evaluate the performance of our proposed algorithm with simulation. Numerical examples show that our proposed algorithm is effective for the power control in cognitive radio networks.

I. I NTRODUCTION Recently, wireless communication technologies such as Wireless LAN, 3G cellular system, and WiMAX are widely utilized. Those wireless communication utilizes frequencies aggressively, and hence it is indispensable to utilize radio frequencies effectively. Cognitive radio is one of the most promising technologies for the effective utilization of wireless resources [1]. In cognitive radio networks, unlicensed (secondary) networks can utilize channels that have been provided for incumbent licensed (primary) networks by avoiding the interference with primary networks. Currently, in order to measure the interference between primary networks and secondary networks, Federal Communications Commission (FCC) has introduced the interference temperature as a metric [2]. When the interference temperature at a measurement point is larger than the interference temperature limit in primary networks, it implies that secondary networks interferes with primary networks. On the other hand, secondary networks should improve the performance of the data transmission. Therefore, each transmitter in secondary networks has to adjust the transmission power carefully so as to both avoid the interference with primary
G. Matsui and K. Sugimoto are with Graduate School of Information Science, Nara Institute of Science and Technology, 8916-5 Takayama, Ikoma, Nara 630-0192, Japan kenji@is.naist.jp Y. Nakamura and S. Wakui are with Institute of Engineering, Tokyo University of Agriculture and Technology, 2-24-16 Nakamachi, Koganei, Tokyo 184-8588, Japan {yukino-n, wakui}@cc.tuat.ac.jp T. Tachibana is with Graduate School of Engineering, University of Fukui, 3-9-1 Bunkyo, Fukui, Fukui 910-8507, Japan

networks and improve quality of service (QoS) of the data transmission. In terms of cognitive radio networks, PID-based power adjustment is also expected to improve QoS of data transmission in secondary networks. On the other hand, the existing PID-based power adjustment algorithm may not avoid the interference with primary networks. This is because the transmission power is determined without considering the interference. Hence, this power adjustment should be extended so as to avoid the interference with primary networks. As far as the authors know, such a power adjustment algorithm that utilizes PID control has not been studied for cognitive radio networks. In this paper, for cognitive radio networks, we propose a PID-based power control algorithm. In this algorithm, at rst, each transmitter in secondary networks receives the feedback information about signal-to-interference plus noise ratio (SINR) from its corresponding receiver. Based on the feedback information, the transmitter derives the transmission power based on PID control in order to satisfy the QoS constraints in secondary networks. Then, the transmitter decides the actual transmission power according to both the derived transmission power and a constraint condition. This constraint condition is utilized to avoid the interference with primary networks, and we also propose an effective update algorithm for the constraint condition. By using both PID control and constraint update algorithm, the improvement of QoS and the avoidance of interference with primary networks can be achieved at the same time. We evaluate the performance of our proposed algorithm with simulation and compare the proposed algorithm with the DPC algorithm [3]. The rest of this paper is organized as follows. Section 2 introduces some related works in terms of the power control for wireless networks. In Section 3, we explain the system model for cognitive radio networks. Then, in Section 4, we explain our proposed power adjustment algorithm based on PID control. We show some numerical examples in Section 5 and present conclusions in Section 6. II. R ELATED W ORK In wireless networks, power control is important issue in order to keep reliable communication between a base station and a user [4]. In order to adjust the transmission power effectively, DPC algorithm has been proposed [3]. This algorithm can be implemented and utilized easily because it does not require complete information on link gains. On the

takuji-t@u-fukui.ac.jp

978-1-4673-0813-7/12/$31.00 2012 IEEE

265

other hand, the DPC algorithm causes the wireless network to become unstable under a small time delay [5]. Therefore, [6] has proposed a power adjustment algorithm based on PID control so that the wireless network becomes stable. In this algorithm, the transmission power is determined by using PID control so as to satisfy the QoS constraints of data transmission. Numerical examples have shown that PID-based method is more effective than DPC in terms of the convergence rate. The PID-based method has also been extended so that a transmitter can utilize multiple PID controllers simultaneously [4]. Moreover, [7] has formulated signal-to-interference ratio (SIR) as a dynamical system and has proposed an optimal power control for mobile users in wireless CDMA networks. The optimal control algorithm has been derived with three cost functions. In [8], sliding mode control theory has been utilized for the power control for ad-hoc wireless networks. For cognitive radio networks, [9] has proposed an auctionbased power control mechanism. In this method, the received power at each receiver is decided by transmitters bid under the constraint of interference temperature. [10] has also proposed another auction-based power control mechanism that can consider QoS in secondary networks in addition to the interference temperature. Those methods are available only when there is only one measurement point in primary networks. Game theory is used to investigate the power control for providing the maximum throughput [11]. In this power control, a tax-based power control game algorithm has been introduced to solve the power allocation optimization problem. [12] has focused on single input multiple output multiple access channels (SIMOMAC). A capped multi-level water lling algorithm and a recursive decoupled power allocation method have been proposed. The performances of those algorithms have been analyzed theoretically. Other power control methods have also been proposed for cognitive radio networks [13], [14]. In [15], the decentralized controller for power adjustment is designed based on the passivity theorem and the ZamesFalb multiplier. [16] has dealt with the power control for the CDMA cellular network system under the condition of unknown channel gains. The common quadratic Lyapunov function approach has been utilized for the stability analysis of wireless network systems with handoff [17]. Thus, for wireless networks including cognitive radio networks, many power control mechanisms have been proposed. Especially, it is well known that PID control is effective for the power control in wireless networks. However, as far as the authors know, a power adjustment method based on PID control has not been studied in detail for cognitive radio networks in spite of its effectiveness. Moreover, our proposed power control scheme is related with the distributed control under input constraints (e.g., [20], [21], [22]). [20] dealt with the consensus problem for doubleintegrator dynamics. [21] and [22] proposed the formation control with the model predictive control. In these works, each agent has a controller and control input is limited similar to our paper. In these works, however, the bound of

g 11
Measurement point 1 M

12
h11

g1 j
g ij

Transmitter 1 M Pair i

h1i i2 h ii

Receiver 1 M

Measurement point j M

Transmitter i
g Ns j

Receiver i
h N si
2 N s

Measurement point N p

g Ns N p

Transmitter N s

h Ns Ns

Receiver N s

Fig. 1.

System model.

the input is constant. In this study, constraints are updated dynamically so as to avoid the interference with primary networks. III. S YSTEM M ODEL Figure 1 shows the system model. In this model, there are Ns (Ns > 0) logical pairs of a transmitter and a receiver in secondary networks. For the ith (i = 1, 2, , Ns ) pair of transmitter i and receiver i, let pi (t) denote the transmission power of transmitter i at time t. When the maximum transmission power is pmax and the minimum one is pmin , pi (t) satises pmin pi (t) pmax . In addition, let hij (hij > 0) be the channel gain from transmitter i to receiver j in secondary networks. The achieved SINR, i (t), for receiver i is given by pi (t)hii i (t) = N . (1)
s

2 pj (t)hji + i (t)

j =1,j =i 2 In (1), i (t) denotes the thermal noise for receiver i at time t. 2 (t) includes the interference from other wireless Note that i

networks including primary networks. To consider effects of the thermal noise, SINR is utilized instead of signal to interference ratio (SIR), which does not include the noise term in the denominator of (1). Now, we focus on the difference ui (t) (pmax + pmin ui (t) pmax pmin ) between the transmission power pi (t) at time t and the transmission power pi (t + 1) at time t + 1. Here, ui (t) is given by ui (t) = pi (t + 1) pi (t). (2) We dene ui (t) as the input for the achieved SINR. From (1) and (2), we have the following dynamical SINR system: hii ui (t)+ i (t + 1) = i (t) + Ii (t) pi (t + 1)hii {Ii (t) Ii (t + 1)} , (3) Ii (t)Ii (t + 1) where
Ns

Ii (t) =
j =1,j =i

2 pj (t)hji + i (t).

(4)

266

In (3), {Ii (t) Ii (t + 1)}/{Ii (t)Ii (t + 1)} can be approximated by zero. This is because the change in the interference from one sample to the next is much smaller than the product of the two consecutive interference values in wireless networks [7]. As a result, we obtain the following SINR system: i (t + 1) = i (t) + hii ui (t). Ii (t) (5)

Pair i in secondary networks Transmitter i Receiver i ui (t ) SINR system i (t ) pi (t ) Plant 2 L p1(t ) p N (t ) s

PID control

zi (t )

Constraint condition update LL

i (t )

pi (t )

Primary networks
y1 (t 1) D1

Note that the SINR system used in [19] is static. On the other hand, the SINR (5) is modelled as a dynamical system so as to adjust the transmission power with PID control. Primary networks consist of Np (Np > 0) measurement points. Let gij (gij > 0) be the channel gain from transmitter i (i = 1, 2, , Ns ) in secondary networks to the j th measurement point (j = 1, 2, , Np ). When Tj is the interference temperature limit for the j th measurement point and B is Boltzmanns constant [18], the maximum interference power tolerance Dj for the j th measurement point is given by Dj = BTj . (6)

1) L y j (t D
j

L L

y N s (t 1) DN s
Measurement N p L

Measurement 1 L

Measurement j L

p1 (t ) p N s (t )

p1 (t ) p N s (t )

p1 (t ) p N s (t )

Fig. 2.

Block diagram of the proposed algorithm.

In cognitive radio networks, transmitter i in secondary networks has to adjust the transmission power by considering the two points: one is the QoS constraints of data transmission in secondary networks and the other is the interference with measurement points in primary networks. In terms of the rst point, SINR i (t) should reach the target SINR. Moreover, in terms of the second point, the total amount of interference power from secondary networks has to be smaller than the maximum interference power tolerance at every measurement points. Therefore, at time t, transmitter i should adjust the transmission power so that pi (t+1) satises the following equations at the next step:
Ns

from receiver i. Here, we assume that the transmission delay of feedback information is unit time due to the simplicity. Note that we can extend our proposed algorithm so as to consider a longer transmission delay. Let pi (t) be the transmission power of transmitter i at time t. We dene zi (t) (pmax + pmin zi (t) pmax pmin ) as the increment of the transmission power. With the feedback information i (t 1), transmitter i derives the desired increment zi (t) by using PID control according to the following equation: zi (t) = KP ei (t 1) + KI xi (t 1) + KD {ei (t 1) (9) ei (t 2)}, where 1 pi (t), i (t 1) xi (t 1) = xi (t 2) + ei (t 1). ei (t 1) = (10) (11)

pi (t + 1)gij +
k=1,k=i

pk (t + 1)gkj Dj .

(7)

Hence, from (2), transmitter i should determine ui (t) as follows:

Ns

{ui (t) + pi (t)}gij +

k=1,k=i

{uk (t) + pk (t)}gkj Dj . (8)

IV. P ROPOSED P OWER C ONTROL A LGORITHM In order to satisfy the QoS constraints and avoid the interference with primary networks simultaneously, we propose a power adjustment algorithm with PID control and dynamic constraint condition. In this section, we explain our power adjustment algorithm. A. Power Control based on PID Control In our power control algorithm, for the above-mentioned system model, transmitter i determines the input ui (t) based on PID control. Figure 2 shows the block diagram of our proposed algorithm. As shown in this gure, at time t, transmitter i receives feedback information about SINR i (t 1)

In (9), KP , KI , and KD denote a proportional gain, an integral gain, and a derivative gain, respectively. Note that the above PID control corresponds to the DPC algorithm in case of the gain KP = 0, KI = 1, and KD = 0 [4]. These equations are the discrete-time version of the proposed PID control due to practical implementation. It is expected that SINR i (t) for each pair i is kept close to the target SINR , improving the performance of secondary networks [6], [3], [4]. B. Constraint Condition Update Algorithm The above power control never guarantees that the interference with primary networks can be avoided. This is because zi (t) is derived by considering only the QoS constraints of data transmission in secondary networks. Therefore, transmitter i has to adjust the transmission power so as to avoid the interference with primary networks by satisfying (8). Here, each transmitter does not know the transmission powers of other transmitters and does not have the information about every link gain. Transmitter i cannot use (8)

267

Start

(t ) < 1
Yes Transmitter i should decrease the transmission power: (t ) pi ( t1 )

No

Transmitter i can increase the transmission power: (t ) pi ( t1 )

receives the information about the maximum interference power tolerance Dj from measurement point j . When Dj is larger than yj (t 1), secondary users do not interfere with measurement point j . With these information transmitter i determines the constraint condition ci (t) at time t. At rst, each transmitter calculates (t) as follows: (t) = Dmin , ymax (t 1) min Dj , yj (t 1). (14)

No Yes Transmitter i should decrease the transmission power: (t ) pi ( t1 )

Yes

No

where Dmin =
j =1,2, ,Np j =1,2, ,Np

(15) (16)

Transmitter i does not change the transmission power: (t ) pi ( t1 )

ymax (t 1) =

max

Transmitter i updates the constraint condition:

ci (t )= (t ) pi ( t1) pi ( t ) (t ) pi ( t1) pi ( t )
if otherwise

If is smaller than one, secondary networks have interfered with primary networks and each transmitter has to decrease his own transmitter power. When is larger than one, each transmitter may increase the transmitted power to improve the performance of secondary networks. Then, the transmitter i investigates the QoS of its own data transmission by deriving i ; i (t) = . i (t 1) (17)

t+1

Fig. 3.

Constraint condition update algorithm.

directly for the power control. Therefore, we introduce a constraint condition into the power adjustment. Now, let ci (t) (pmax + pmin ci (t) pmax pmin ) denote the dynamic constraint condition for the desired increment zi (t) of transmission power at time t. The constraint condition ci (t) is determined so that the interference with primary networks can be avoided. After transmitter i derives zi (t) from (9), it compares zi (t) with ci (t). Transmitter i decides the actual increment ui (t), that is dened in (2), according to the following equation: ui (t) = min{zi (t), ci (t)}. (12)

When i (t) is larger than one, transmitter i should increase the transmission power so as to increase SINR. If i (t) is equal to or smaller than one, each transmitter should decrease the transmission power in order to increase SINRs of other data transmissions. From (t) and i (t), transmitter i updates the constraint condition ci (t) according to the following equations: ci (t) = i (t)(t)pi (t 1) pi (t), (t)pi (t 1) pi (t), i (t) 1, i (t) > 1. (18)

When ci (t) is large (small), the QoS for transmitter i tends to be high (low) due to a large (small) transmission power. Therefore, ci (t) has to be determined so as to improve the performance of secondary networks while avoiding the interference with primary networks. Hence, we propose an effective update algorithm for the constraint condition. Our proposed algorithm is shown in Fig. 3. Let yj (t) (j = 1, 2, , Np ) be the observed interference power at measurement point j at time t (see (8));
Ns

Dmin In these equations, (t)pi (t 1) = ymax (t1) pi (t 1) means that the interferences are likely to be avoided at all measurement points. Moreover, when i (t) is equal to or smaller than one, transmitter i tries to decrease the transmission power with i (t)pi (t 1) so that i (t) becomes equal to and j (t) (j = i) is increased. When i (t) is larger than one, transmitter i would like to increase the transmission power with i (t)pi (t 1) for improving the QoS of its own data transmission. However, in (18), we do not multiply pi (t 1) by i (t) because a large i (t) may occur the interference with primary networks.

V. N UMERICAL E XAMPLES In this section, we evaluate with simulation the performance of our proposed algorithm over a 1000 [m] 1000 [m] square area. In this area, as shown in Fig. 4, there is a secondary network with six transmitter-receiver pairs (Ns = 6). Those pairs are randomly distributed over the area. In addition, there are three measurement points that are also randomly distributed over the area (Np = 3). In the following, a unit time is set to 0.01 [s], and simulation time is 50 [s].

yj (t) =
k=1

pk (t)gkj .

(13)

Each transmitter receives information about yj (t 1) at time t, as shown in Fig. 2. Here, we assume that the transmission delay of feedback information from each measurement point is unit time due to the simplicity. Every transmitter also

268

1000 Transmitter

x 10 7

14

800
Position y [m]

Receiver

= 5 .0 10 14 Measurement point 3

DPC PI D

600

1 6

Inte rference power

Measurement point

6 5 4 3 2

400 4 5 200 2 3 0 0 200 400 600 Position x [m] 800 1000 3

Measurement point 2

1 0 0

Measurement point 1
10 20 30 40 50

Fig. 4.

Simulation scenario where Ns = 6 and Np = 3. Fig. 5.

Ti me [s]

For the secondary network, the target SINR is set to 2 (t) is equal to regardless of receiver i, and 0.033. Noise i is independent and identically distributed (i.i.d.) random variable with a uniform distribution on [1.0 1014 , 2.0 1014 ]. When the distance between transmitter i and receiver 4 j is dij [m], the channel gain hij is given by hij = d ij [11]. The minimum transmission power and the maximum one are given by pmin = 0 [W] and pmax = 0.01 [W], respectively. For the primary network, we assume that the maximum interference power tolerance Dj is equal to D for each measurement point j . In addition, when the distance between transmitter i and measurement point j is dij [m], the channel 4 gain gij is also given by gij = d ij . Under this situation, each transmitter in the secondary network utilizes our proposed algorithm. Each transmitter adjusts the transmission power by performing PID control at every 10 [ms]. In PID-based algorithm, KP , KI , and KD are set to 0.46, 0.03, and 0.01, respectively. A. Time Responses of Interference Temperature and SINR First, we investigate how interference temperature at measurement points and SINR change over time by using our proposed algorithm. Here, the maximum interference power tolerance is set to D = 5.0 1014 . Figure 5 shows the interference power (i.e., left-hand side of (7)) at each measurement point against time, and Fig. 6 shows SINR against time. From Fig. 5, we can nd that each transmitter can avoid the interference with primary network. This implies that the constraint condition is effective for the interference avoidance. In Fig. 6, SINR is smaller than at each receiver. This is because the transmission power becomes small due to a small constraint condition. Nevertheless, each transmitter tries to improve the performance of secondary network. B. Performance Comparison in Terms of SINR and Transmission Power Then, we investigate how the maximum interference power tolerance D affects the performances of our proposed algorithm. Here, D changes from 1.0 1014 to 1.0 1013. In this performance comparison, we calculate the variance of

Interference temperature vs. time in case of D = 5.0 1014 .

DPC PI D

0.05

receiver 3 SINR
0.033

0 0 0.05

10

20

30

40

50

Time [s]
receiver 5 SINR
0.033

PID

DPC

0 0

10

20

30

40

50

Time [s]
Fig. 6. SINRs of receivers 3 and 5 vs. time in case of D = 5.0 1014 .

SINR for the ith transmitter-receiver pair from 0 [s] to 50 [s] as follows: Vi = 1 5001
5000 t=0

{i (t) }2 .

(19)

Then, we derive the average of variance over all pairs as Ns 1 i=1 Vi . When the average of the variance is small, Ns our proposed algorithm can improve the performance of secondary network. In addition, we calculate the average transmission power of transmitter i from 0 [s] to 50 [s] as follows: Pi = 1 5001
5000

pi (t).
t=0

(20)

Then, we derive the average of the transmitter over all pairs Ns 1 as N i=1 Pi . When the average of the transmission power s is small, our proposed algorithm can save the power. Figure 7 shows the average of the deviation of SINR. From this gure, we can see that the average of the variance of our proposed algorithm is smaller than that of the DPC algorithm. Figure 8 shows the average of the transmission power. From this gure, we can nd the average of the

269

10

x 10

R EFERENCES
[1] J. Mitola, Cognitive Radio: Making Software Radios More Personal, IEEE Personal Communications, vol. 6, no. 4, pp. 13-18, Aug. 1999. [2] Federal Communication Commission, Spectrum Policy Task Force, Rep. ET Docket, no. 02-135, Nov. 2002. [3] G.J. Foschini and Z. Miljanic, A Simple Distributed Autonomous Power Control Algorithm and its Convergence, IEEE Trans. Veh. Technol., vol. 42, no. 4, pp. 641-646, Nov. 1993. [4] A. Paul, A. Member, M.G. Safonov, and U. Mitra, Adaptive Power Control for Wireless Networks Using Multiple Controllers and Switching, IEEE Trans. Neural Netw., vol. 16, no. 5, pp. 1212-1218, Sept. 2005. [5] F. Gunnarsson, J. Blom, and F. Gustafsson, Power Control in Cellular Systems subject to Constraints and Time Delays, in Proc. IEEE Globecom 1998, pp. 36453650, Nov. 1998. [6] F. Gunnarsson, F. Gustafsson, and J. Blom, Pole Placement Design of Power Control Algorithms, in Proc. IEEE 49th Vehicular Technology Conference, pp. 21492153, May 1999. [7] S. Koskie and Z. Gajic, Optimal SIR-based Power Control Strategies for Wireless CDMA Networks, International Journal of Information and Systems Sciences, vol. 4, no. 2, pp. 204-218, Jan. 2008. [8] M.R. Cistelecan and D.C. Popescu, Using Sliding Mode Control Theory for Radio Resource Management in Wireless Systems, in Proc. 9th IEEE International Workshop on Variable Structure Systems 06, pp. 23-28, June 2006. [9] J. Huang, R. Berry, and M.L. Honig, Auction-based Spectrum Sharing, ACM/Springer Mobile Networks and Applications Journal (MONET), vol. 11, no. 2, pp. 405-408, Apr. 2006. [10] Y. Xing, C.N. Mathur, and M.A. Haleem, Dynamic Spectrum Access with QoS and Interference Temperature Constraints, IEEE Transactions on Mobile Computing, vol. 6, no. 4, pp. 423 - 433, Apr. 2007. [11] H. Li, Y. Gai, Z. He, K. Niu, and W. Wu, Optimal Power Control Game Algorithm for Cognitive Radio Networks with Multiple Interference Temperature Limits, in Proc. VTC-2008 Spring, May 2009. [12] L. Zhang, Y.C. Liang, and Y. Xin, Joint Beamforming and Power Allocation for Multiple Access Channels in Cognitive Radio Networks, IEEE Journal on Selected Areas in Communications, vol. 26, no. 1, pp. 38-51, Jan. 2008. [13] A.T. Hoang, Y.C. Liang, and M.H. Islam, Power Control and Channel Allocation in Cognitive Radio Networks with Primary Users Cooperation, IEEE Trans. Mobile Computing, vol. 9, no. 3, pp. 348-360, Mar. 2010. [14] K. Hamdi, W. Zhang, and K.B. Letaief, Power Control in Cognitive Radio Systems Based on Spectrum Sensing Side Information, in Proc. IEEE International Conference on Communications 2007, pp. 51615165, June 2007. [15] V.V. Kulkarni, M.V. Kothare, and M.G. Safonov, Decentralized Dynamic Nonlinear Controllers to Minimize Transmit Power in Cellular Networks, Part I, Systems & Control Letters, vol. 59, no. 5, 294298, May 2010. [16] S. Jagannathan, M. Zawodniok, and Q. Shang, Distributed Power Control for Cellular Networks in the Presence of Channel Uncertainties, IEEE Transactions on Wireless Communications, vol. 5, no. 3, pp. 540549, Mar. 2006. [17] A. Paul, M. Akar, U. Mitra, and M.G. Safonov, A Switched System Model for Stability Analysis of Distributed Power Control Algorithms for Cellular Communications, in Proc. 2004 American Control Conference, pp. 16551660, June 2004. [18] W. Wang, T. Peng, and W. Wang, Optimal Power Control under Interference Temperature Constraints in Cognitive Radio Network, in Proc. IEEE WCNC 2007, pp. 116120, Mar. 2007. [19] S. Koskie and Z. Gajic, A Nash Game Algorithm for SIR-based Power Control in 3G Wireless CDMA Networks, IEEE/ACM Transactions on Networking, vol. 13, no. 5, pp. 1017-1026, Oct. 2005. [20] W. Ren, On Consensus Algorithms for Double-Integrator Dynamics, IEEE Transactions on Automatic Control, vol. 53, no. 6, pp. 1503 1509, July 2008. [21] T. Yang, Z. Liu, H. Chen, and R. Pei, Distributed Robust Control of Multiple Mobile Robots Formations via Moving Horizon Strategy, in Proc. 2006 American Control Conference, pp. 28382843, June 2006. [22] W. B. Dunbar and R. M. Murray, Distributed Receding Horizon Control for Multi-vehicle Formation Stabilization, Automatica, vol. 42, no. 4, pp. 549558, Apr. 2004.

Average of the variance of SINR

DPC PID 1 2 3 4 5 6 7 8 9 x 10 10
14

Fig. 7.
x 10 5
6

Average deviation of vs. time.

Average of transmission power

DPC PID 1 2 3 4 5 6 7 8 9 x 10 10
14

Fig. 8.

Average transmission power vs. time.

transmission power of our proposed algorithm is almost the same as that of the DPC algorithm. Therefore our proposed power control algorithm is more effective than DPC algorithm. VI. C ONCLUSION AND F UTURE W ORK In this paper, we proposed a power adjustment algorithm based on PID control and dynamic constraint condition for cognitive radio networks. Numerical examples showed that each transmitter can control his own transmission power so as to improve the performance of secondary networks while avoiding the interference by using our proposed algorithm. Moreover, we found that our proposed algorithm is more effective than a conventional distributed power control in cognitive radio networks. In the research of distributed control systems with input constraints, the stability of the global system has been analyzed [20], [21], [22]. In our future work, we will prove the stability of our proposed control system. Moreover, we will use the quadratic programming instead of PID control in order to improve the performance. ACKNOWLEDGEMENT This research was supported in part by the Kinki Mobile Radio Center Foundation.

270