method for power control is proposed for broadband, packet-switched TDMA wireless networks. By observing the temporal correlation of cochannel interference when transmitters can send data contiguously, a Kalman filter is used to predict interfence power in
lter method yields observable improvement even when amessage consists of a few packets on average. Theseresults are discussed as follows.5. PERFORMANCE STUDY
Given the complexity of the system and controlmethod, we choose to use simulation techniques to studythe effectiveness of the Kalman- lter method. In thefollowing, we rst present the details of the simulationmodel and then discuss the performance results.5.1 Simulation Model
We simulate the cell layout and interleaved channelassignment (ICA) [WL98] in Figure 1. A total of 19 cellsare simulated. Each cell is divided into 4 sectors, each ofwhich is served by a base station antenna located at thecenter of the cell. The beamwidth of each base stationantenna is 60o, while terminals have omni-directionalantennas. The radiation pattern for the base stationantenna is assumed to be a parabolic shape; that is, a 3dBdrop occurs at the beamwidth half angle and any directionbeyond a threshold angle in clockwise or anti-clockwisedirection suffers a given, xed attenuation relative to thegain at the front direction, which is called thefront-to-back(FTB) ratio. Thus, given the beamwidth and the FTBratio, the parabolic-shape antenna pattern can be fullyspeci ed. For the 60obase station antenna with 20dBFTB ratio, this pattern yields a 3dB drop at the 30oanglein clockwise or anti-clockwise direction from the frontdirection, the threshold angle is 77. 5oand the antenna gainat the front direction is 9.5dBi. The ICA (static) scheme inFigure 1 is considered where the frequency reuse factor is2/8 (i.e., reuse in every 2 cells or 8 sectors).
Each radio link between a terminal and its base stationis characterized by a path-loss model with an exponent of4 [R96] and lognormal shadow fading with a standarddeviation of 8dB. Fast fading is not considered in thisstudy. Cell radius is assumed to be 1 Km and the path lossat 100m from the cell center is -70dB. Thermal noisepower is xed and equal to -115dBm (to approximate a useof 1MHz channel bandwidth).
To fully consider the effects of shadow fading and theantenna pattern, terminals are rst placed randomly at±67. 5ofrom the front direction of the base station antennaand as much as 1. 25 times the cell radius from the centerof each cell. Then, each terminal selects the base stationthat provides the strongest signal power. The process isrepeated until each sector serves 500 terminals. Toprovide accurate results, only statistics in the middle cell inFigure 1 are collected and reported below. In addition,each simulation is repeated with 5 different sets of randomseeds (e.g., for populating terminals and selecting shadowfading) to ensure correctness and results presented beloware aggregated results of all ve sets.
Message length is assumed to have a discrete form ofPareto distribution. We use a Pareto distribution because ithas been shown to be appropriate for modeling IP traf c[WTSW97] and used by [U97] and [CS97] to study
wireless network performance. More precisely, each timea terminal transmits a message, the number ofislots(packets) used in the message transmission is characterizedby the following cumulative distribution function:
k _ Hi=1 _ i
α
fori≥k∈Z+andα >1.(14)
wherekandαare given parameters. Then, the probabilitythat an arbitrary message consists ofipackets is given by
k k _ _ ____ hi= _ i i+1
α
α
fori≥k∈Z+.
(15)
After some algebra, the average message length is
L=k+
Σi=k+1
∞
k
_ . _ i
α
(16)
It is known that the Pareto distribution has an in nitevariance ifα ≤2. In nite variance should be avoided,otherwise our simulation model cannot reach a steady stateand results may not have statistical signi cance. Thus, to
L_ guarantee nite variance, for a givenL, we setk= _ 2
to ensureα >2 where y is the smallest integer greaterthan or equal toy. Using thiskvalue,αcan be solvedfrom (16) to match the given average message lengthL.Two adjustable parametersWandηfor (9) and (10)are set to be 30 and 0.5, respectively. As discussed earlier,W=30 is appropriate asL≤10 in our study, whereasη =0. 5 is appropriate because ifηis too small, the KalmangainKn’s by (5) will be very close to 1, thus making thepredicted interference level too close to the lastmeasurement. Our numerical experiments reveal that theKalman- lter method gives similar results forW=50 , 80andηranging from 0.25 to 1.
For convenience, our simulation model assumes thatterminals in all cells are synchronized at the slot boundaryfor transmission. (This assumption can be relaxed inpractice with some minor performance degradation.)Furthermore, we assume 100% traf c load in this study.That is, there are always terminals ready for transmissionin co-channel sectors. Thus, after a terminal transmits amessage with a random length according to (14), the basestation immediately schedules another randomly chosenterminal in the same sector to start a new transmission inthe next time slot. To help us study the dynamic range forthe power control schemes, the model assumes no limit onthe actual transmission power.
Unless stated otherwise, the model also assumes thatinterference power in one time slot can be measured andused to determine the transmission power for the next slot.Nevertheless, a study of performance impacts due to delayin providing the transmission power to the transmittingterminal is also reported in the following.5.2 Performance Results and Discussions
Figure 2 compares the SINR performance for theKalman- lter method with that for no, full and optimal
(0); //initialize with a measurement x_est_last = z_real + frand()*0.09; float sum_error_kalman = 0; float sum_error_measure = 0; for (...
% [x, V, VV, loglik] = kalman_filter(y, A, C, Q, R, init_x, init_V, ...) % % INPUTS: % y(:,t) - the observation at time t % ...
源于1957年他在哥伦比亚大学的一篇博士 论文和1960年发表的论文《A New Approach to Linear Filtering and Prediction Problems》,60年代Kalman 把状态空间模型引入滤波...
y〔n〕=Cx〔n〕 为 A = 〔1.1269-0.49400.1129 1.0000 0 B = 〔-0.3832 0.5919 0.5191〕; C = 〔1 0 0〕; 我们的目标是设计 Kalman 滤波...
卡尔曼全名 Rudolf Emil Kalman,匈牙利数学家,1930 年出生于匈牙利首都布达佩斯。...4. 简单例子 (A Simple Example) 这里我们结合第二第三节, 举一个非常简单...
(cm == 'A') mtype = 2; else error('Unkown measurement model.'); ...(sd,T.nmeas); % Kalman gain % Initialize filter L_init = L_max/2.0...
Kalman Filter介绍_数学_自然科学_专业资料。卡尔曼滤波(kalman filter)最简单...4 简单例子(A Simple Example) 这里我们结合第二第三节,举一个非常简单的例子...
(A,[B B],C,0,-1,'inputname',{'u' 'w'},'outputname','y'); Q = 1; R = 1; [kalmf,L,P,M] = kalman(Plant,Q,R); kalmf = kalmf(...
Kalman 滤波 在学习卡尔曼滤波器之前,首先看看为什么叫“卡尔曼”。跟其他著名...(k-1k-1)对应的 covariance,A’表示 A 的转置矩阵,Q 是系统过程的 ...
Kalman滤波中文版 17页 1下载券 kalman滤波的C与C++实现 10页 1下载券 ... Kalman滤波的公式推导...
热门文档