T. Ekman, Prediction of Mobile Radio Channels. Modeling and Design. PhD Thesis, Signals and Systems, Uppsala University, Sweden, 2002. Online: www.signal.uu.se/Publications/abstracts/a023.html  Home/Search   Document Not in Database   Summary   Related Articles   Check

....over such horizons is difficult, and has only recently begun to be investigated. We have developed predictors that use channel tap estimates as regressors to predict the power of all significant channel taps. The predictors have been validated on an extensive set of broadband channel measurements [10, 11], see Figure 1 for an illustration. A recently developed unbiased quadratic predictor [3] provides the so far best performance, see Figure 2. We have also obtained a good theoretical understanding of the limits of attainable performance in different propagation environments. 3 The Link and MAC ....

....The prediction horizon is 5.5ms. The prediction (solid) is overlaid on a least squares estimate of a true tap (thin line) Measured channel sounding data is used. Note that it is possible to predict beyond a fading dip, so that the transmission can be planned accordingly. For details, see [3] [10], 11] 0.2 0.4 0.6 0.8 Average power prediction perfomance Prediction range [wavelengths] NMSE Unbiased pred. Biased pred. Last samp. Average pred. Figure 2: Power prediction on 5 MHz channels at 1880MHz. The normalized MSE is evaluated as a function of the prediction range, ....

T. Ekman, Prediction of Mobile Radio Channels. Licentiate Thesis, Signals and Systems, Uppsala University, Sweden, Dec. 2000.

....overprotective channel coding, and information that the wireless terminal is unable to present. The statement also implies that the applications should be able to adjust their requirements on the requested services from the communication system. Recent work by Torbj orn Ekman and others [15, 16, 56], suggests that predictions of channel quality can be achieved for several milliseconds ahead in time. In this thesis we try to answer the question of how accurate, or, how long predictions we need in order to perform ecient resource allocations through scheduling. An alternative approach to a ....

....2. 1 The Nature of Fading If we regard the received radio waves as being composed of simple sinusoids with varying amplitudes, phases, and frequencies, and assume that all the parameters can be estimated, then the received power can be found as a function of the distance to the radio source [1, 15, 51]. The fading occurs when many sinusoids interfere, adding up to a varying resulting power at each point in space. 2.1.1 A Multipath Fading Experience When sitting at the edge of a swimming pool, splashing your feet in the water, you immediately observe waves propagating out, in a circle, from ....

[Article contains additional citation context not shown here]

T. Ekman. Prediction of Mobile Radio Channels. Licentiate Thesis, Uppsala University, Dec. 2000.

.... out in the Wireless IP (WIP) project within the Swedish research program for Personal Computing and Communications (PCC) 2] Our results this far show that a high gain in spectral efficiency can be obtained using channel prediction and scheduling algorithms [7, 14] Through channel prediction [5], we obtain estimates of future transmission quality over the channels between a base station and several mobile terminals. These estimates are fundamental in the following step, the media access scheduling. The scheduler s task is to match the offered Internet traffic to the different channels, ....

....type of simulation requires excessive calculations when a realistic outcome is desired. We use methods that involve ray tracing of individual paths of the radio signals, in order to quantify the interference at specific points in space (the points where the receiving mobile terminals are situated) [5]. We have previously been using very short recordings of channel sounding measurments, that is, a kind of pinging from a base station that is recorded by a mobile receiver. These measurments only allowed us to run the simulations for about # # # , which is much too short to see the ....

T. Ekman. Prediction of Mobile Radio Channels. Licentiate Thesis, Uppsala University, Dec. 2000.

.... are possible by making channel state information available to the system and the transmitters [22] 23] 24] This however requires that channel prediction algorithms that can predict the channel impulse response well into the future can be designed and that these have reasonable good performance [25]. Furthermore, adaptive coding modulation schemes, fast scheduling algorithms, and multiple access schemes that can use the information (predicted channel impulse response and the accuracy of it) provided by the channel prediction algorithm must be developed. A feedback channel is needed in the ....

T. Ekman, Prediction of Mobile Radio Channels. Lic. Thesis, Uppsala University, Sweden, Dec. 2000.

....and Coding with ARQ) and one plain uncoded AMS. Both are combined with a time slot scheduler supplied with channel quality predictions. Basically, we assume that the channel quality for each radio link, here in terms of the SNR, can be predicted for a time interval of about 10 ms into the future [8, 9], and that the predictions are accessible at the link layer. Based on the predicted values, the link layer preliminary selects a Modulation and Coding Scheme (MCS) for each tratfic flow which satisfies the QoS requirements and provides high throughput. The scheduler uses the information about the ....

T. Ekman, "Prediction of Mobile Radio Channels," Phd thesis, Signals and Systems, Uppsala University, Sweden, Oct. 2002.

.... are possible by making channel state information available to the system and the transmitters [21] 22] This however requires that channel prediction algorithms that can predict the channel impulse response well into the future can be designed and that these have reasonable good performance [23]. Furthermore, adaptive coding modulation schemes, fast scheduling algorithms, and multiple access schemes that can use the information (predicted channel impulse response and the accuracy of it) provided by the channel prediction algorithm must be developed. A feedback channel is needed in the ....

T. Ekman, Prediction of Mobile Radio Channels. Lic. Thesis, Uppsala University, Sweden, Dec. 2000.

....and high speed services. In order to perform a simplified analysis, see Section 3 below, we specialize to the case where users are given exclusive access to the broadband data channel, and where a single 200kHz narrowband channel is used. A Channel Prediction By predicting the short term fading [6, 7], the transmission load among the users can be distributed in an optimal way, while fulfilling certain specified constraints on throughput and delay. From extensive investigations on broadband measurement data, we have found that the received signal power distribution in the time domain as well ....

.... From extensive investigations on broadband measurement data, we have found that the received signal power distribution in the time domain as well as in the frequency domain can be predicted with reasonable accuracy for a time interval corresponding to a movement of up to half a carrier wavelength [6, 7]. 2 Half a wavelength at 1900MHz would correspond to about 10 of our proposed 0.5ms slots at vehicular speeds of about km h. In such a situation we may allocate transmission loads among the users over the corresponding scheduling horizon. Interference reducing schemes described in Sections B ....

[Article contains additional citation context not shown here]

T. Ekman, Prediction of Mobile Radio Channels. Licentiate Thesis, Signals and Systems, Uppsala University, Sweden, Dec. 2000.

....a cell can be predicted a number of milliseconds ahead, then it would be possible to distribute the transmission load among the users in an optimal way, while fulfilling certain specified constraints on throughput and delays. The achievable prediction performance depends on the available bandwidth [2, 3]. Here the discussion of predictors will assume a 5 MHz bandwidth if not stated differently. From extensive investigations on broadband measurement data, we have found that the received signal power can be predicted with reasonable accuracy for a time interval corresponding to a movement of up to ....

....predictors will assume a 5 MHz bandwidth if not stated differently. From extensive investigations on broadband measurement data, we have found that the received signal power can be predicted with reasonable accuracy for a time interval corresponding to a movement of up to half a carrier wavelength [2, 3] . In a WCDMA context for HSDPA [4, 5] half a wavelength would correspond to 9 slots (# # ms at ## km h) We may thus allocate transmission loads among the users over these nine slots based on their predicted channel qualities. In HSDPA one user at a time will in general use the whole bandwidth. ....

[Article contains additional citation context not shown here]

T. Ekman, Prediction of Mobile Radio Channels. Licentiate Thesis, Signals and Systems, Uppsala University, Sweden, Dec. 2000.

....while satisfying the required QoS requirements [4 8] In this work we propose a HARQ II AMS combined with a time slot scheduler supplied by channel predictions. Basically, we assume that the channel quality for each radio link can be predicted for a time interval of about 10 ms into the future [9,10], and that these predictions are accessible by the link layer. Based on the predicted values, the HARQ II AMS preliminary selects a Modulation and Coding Scheme (MCS) for each user which satisfies the BER requirement and provides high throughput. The scheduler uses the information about the ....

T. Ekman, Prediction of Mobile Radio Channels, Technical licentiate thesis, Dep. of Signals and Systems, Uppsala University of Technology, Uppsala, Sweden, Dec. 2000.

....a cell can be predicted a number of milliseconds ahead, then it would be possible to distribute the transmission load among the users in an optimal way, while fulfilling certain specified constraints on throughput and delays. The achievable prediction performance depends on the available bandwidth [2, 3]. Here the discussion of predictors will assume a 5 MHz bandwidth if not stated differently. From extensive investigations on broadband measurement data, we have found that the received signal power can be predicted with reasonable accuracy for a time interval corresponding to a movement of up to ....

....predictors will assume a 5 MHz bandwidth if not stated differently. From extensive investigations on broadband measurement data, we have found that the received signal power can be predicted with reasonable accuracy for a time interval corresponding to a movement of up to half a carrier wavelength [2, 3] . In a WCDMA context for HSDPA [4, 5] half a wavelength would correspond to 9 slots ( ms at km h) We may thus allocate transmission loads among the users over these nine slots based on their predicted channel qualities. In HSDPA one user at a time will in general use the whole ....

[Article contains additional citation context not shown here]

T. Ekman, Prediction of Mobile Radio Channels. Licentiate Thesis, Signals and Systems, Uppsala University, Sweden, Dec. 2000.

....while satisfying the required QoS for different users [4 7] In this work we propose a HARQ II AMS combined with a time slot scheduler supplied by channel predictions. Basically, we assume the channel quality for each radio link can be predicted for a time interval of about 10 ms into future [8, 9], and that they are accessible by the link layer. Based on the predicted values, the HARQII AMS preliminary selects a Modulation and Coding Scheme (MCS) for each user which satisfies the QoS and provides high throughput. The scheduler uses the information about the individual data streams, along ....

T. Ekman, Prediction of Mobile Radio Channels, Licentiate thesis, Department of Signals and Systems, Uppsala University of Technology, Dec. 2000.

....When evaluating these and other issues, we need assumptions about the wireless links. Within the Wireless IP project [17] we assume multiple users who share fading wideband links in a coordinated way. The resource sharing and optimization is based on predictions of the channel quality [10] and uses adaptive modulation and scheduling [11] as outlined briefly in the next section. Thus, information on the link quality is produced and used at lower layers, and it could be used also by higher layers, if desired. Our long term aim will be to cooptimize the adaptive modulation, ....

T. Ekman "Prediction of Mobile Radio Channels ", Licentiate Thesis, Signals and Systems, Uppsala University, December 2000.

....These properties are exploited in [24] where a continuous time predictor for the Jakes channel is presented. The use of quadratic predictors [25] or MARS model based predictors [26] for the complex tap results in predictors with poor generalization properties outside the estimation interval [27]. Channel prediction on measured channels Prediction results for the Jakes channel or channels modeled as a low number of sinusoids should not be taken as an evidence for high predictability of real mobile radio channels. In practice the dynamics of the taps are not perfectly bandlimited, due to ....

....will dominate the model (2.119) Thus, even if the correct model is ARMA, an AR model might o#er several advantages and a su#ciently good fit to the channel dynamics. As it generally is harder to estimate zeros as compared to poles, the AR model has been preferred in many channel predictors [22] [27]. In this thesis we will consider both AR and ARMA models for the channel, in the design of the channel predictors. Time varying models When the transceiver moves, the angle towards any given cluster changes. This result in a corresponding change of the position of the pole in the AR1 model for ....

[Article contains additional citation context not shown here]

T. Ekman, Prediction of mobile radio channels, PhD thesis, Uppsala University, Uppsala, Sweden, 2001.

....obtained by Ericsson Research in urban and suburban areas, at the carrier frequency 1880 MHz and with a 6.4 MHz baseband sampling rate, have been utilized. The sampling rate of the channel is 9.14 kHz whereas the highest Doppler frequency is about 160 Hz. Thus we have a highly oversampled channel [6]. The channel can be modeled as a time varying FIRfilter, with taps denoted by h k (t) Let n denote a discrete time index at the symbol rate. For each measurement location, the time varying complex valued channel taps h k (n) in s(n) h k (n)u(n k) w(n) 1) were estimated. Here, s(n) is ....

....error v k (t) will be close to white when considering consecutive snapshots. For a given smoothing lag and filter order, a FIR Wiener smoother is parameterized by the SNR and the Doppler frequency. These entities can be estimated from the power delay profile and the average Doppler spectrum [6]. In the measurements at hand, the Doppler frequency normalized by the channel sampling frequency is on the order of 10 mHz. This corresponds to 100 samples per traveled distance measured in wavelengths. Smoothing (use of future channel samples) provides much better noise reduction performance ....

T. Ekman , "Prediction of Mobile Radio Channels," Licentiate thesis, Uppsala University, Dec. 2000.

....obtained by Ericsson Research in urban and suburban areas, at the carrier frequency 1880 MHz and with a 6.4 MHz baseband sampling rate, have been utilized. The sampling rate of the channel is 9.14 kHz whereas the highest Doppler frequency is about 160 Hz. Thus we have a highly oversampled channel [6]. The channel can be modeled as a time varying FIR filter, with taps denoted by h k (t) Letndenote a discrete time index at the symbol rate. For each measurement location, the timevarying complex valued channel taps h k (n) in s(n) h k (n)u(n k) w(n) 1) were estimated. Here, s(n) is ....

....error v k (t) will be close to white when considering consecutive snapshots. For a given smoothing lag and filter order, a FIR Wiener smoother is parameterized by the SNR and the Doppler frequency. These entities can be estimated from the power delay profile and the average Doppler spectrum [6]. In the measurements at hand, the Doppler frequency normalized by the channel sampling frequency is on the order of 10 mHz. This corresponds to 100 samples per traveled distance measured in wavelengths. Smoothing (use of future channel samples) provides much better noise reduction performance ....

T. Ekman , "Prediction of Mobile Radio Channels," Licentiate thesis, Uppsala University, Dec. 2000.

....a single outdated fading estimate is proposed. In [19 21] adaptive modulation schemes based on accurate long rang CSI prediction is investigated. Here, the proposed system utilizes an unbiased quadratic regression of historic noisy channel estimates to predict the signal power at the receiver [22, 23]. For this type of predictor, there exists a statistical model for the prediction error for Rayleigh fading channels. The model provides the mean, variance and probability density function of the errors. This statistical model enables an analytical optimization of the rate adaptation scheme. It ....

.... t) rg(L (K t) T. tt) The use of (t0) in (9) gives E(p ln ) rg. The bias compensation in (9) will reduce the total prediction MSE. It also provides superior performance compared to the use of linear power predictors that are based on channel power samples (lyl 2) as regressors [23] 2. The miminum MSEs of the channel gain prediction error (i.e. e = g l ) and the power prediction error (i.e. ep = p Pl ) are given by H 1 Iff e2c T g 7 g rgR rg (12) and 2 2 H 1 2 = IrgR rail , 7 ep (13) respectively. Thus, the optimum quadratic power predictor can be ....

[Article contains additional citation context not shown here]

T. Ekman, "Prediction of Mobile Radio Channels," Phd thesis, Signals and Systems, Uppsala University, Sweden, Oct. 2002.

....[2] 3] 7] 8] Our proposed scheme is based on fast adaptive modulation and scheduling of the IP traffic that adapts to the short term fading. We use predictions of the future channel quality for all active mobile terminals and are developing predictor algorithms for this purpose [9] 10] [11]. See Figure 1 for an illustration. 20 40 60 80 100 120 140 40 30 20 10 0 10 Time (ms) Rel. ampl. dB) Figure 1: Typical long term prediction result for one channel tap in a channel of bandwidth 5MHz at 1900 MHz. The prediction horizon is 5.5ms. The solid line is the prediction, ....

T. Ekman, Prediction of Mobile Radio Channels. Licentiate Thesis, Signals and Systems, Uppsala University, December 2000.

No context found.

T. Ekman, Prediction of Mobile Radio Channels. Modeling and Design. PhD Thesis, Signals and Systems, Uppsala University, Sweden, 2002. Online: www.signal.uu.se/Publications/abstracts/a023.html

No context found.

T. Ekman, Prediction of Mobile Radio Channels. Licentiate Thesis, Signals and Systems, Uppsala University, Sweden, Dec. 2000.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC