Communications Services Is Constantly Growing Computer Science Essay

Published: Last Edited:

This essay has been submitted by a student. This is not an example of the work written by our professional essay writers.

The IEEE 802.11p or Dedicated Short Range Communication, DSRC is the communication standard for vehicle-to-vehicle and vehicle-to-infrastructure V2I communications but this standard has weak performance in multipath environments since IEEE 802.11p is very similar to the IEEE 802.11a standards. It is shown that conventional estimation methods do not perform sufficiently well because of factors like the high relative speed of vehicles and the delay spread of the channel. In this paper, we develop estimation method (EM-SBL) that uses for sparse channel structure and combine it with feedback estimators to closely track the V2V channel dynamics. The channel model used in simulation is referencing HIPERLAN/2 Model C.

Keywords: 802.11p, Channel Estimation, Data Decision Feedback, EM-SBL


The need for communications services is constantly growing. Nowadays mobile devices are present almost everywhere and thus enable a number of future vehicular applications such as safety, information about real-time traffic, road conditions or the state of other vehicles.

To perform these applications, it is necessary to have a physical-layer protocol which is able to provide reliable V2V communication in a wide range variety of driving environments. DSRC (or IEEE 802.11p) is a short-range wireless communication technology [2] to work in the 5.9 GHz band. The IEEE 802.11p exploits orthogonal division multiplexing (OFDM) transmission with 10 MHz channel spacing. Listed in Table.1 are the parameters employed in IEEE 802.11p. It actually adopts the same PHY defined for 802.11a, except for channel bandwidth being used.

The channel characteristics of V2V channels are fundamentally different from indoor channels because of mobility or large stationary reflectors (such as buildings, hills, trees and signs). The impulse response of V2V multipath channel can be modeled as a tapped delay line filter. Although the number of non-zero components is not small enough to categorize the channel as sparse, HIPERLAN/2 (Model C) is to be available for sparsity analyze. In this situation, compressed sensing (CS) has emerged as one such frame work, in which a K-sparse (only K entries of an L length vector are non-zero) high dimensional signal can be recovered from a small number of observations [3]. In sparse channel estimation problems, the second order statistics of the channel are assumed to be unknown. Thus, employing a Bayesian perspective, the channel estimation problem can be rephrased by letting where is a deterministic unknown hyperparameter. In particular, Sparse Bayesian Learning (SBL) techniques are applied to jointly estimate the sparse channel and its second order statistics.

Although the comb-type pilot methods are satisfactory in fast fading channels, the pilot organization of IEEE 802.11p standard does not perform well in urban environments, where multipath fading is severe and thus delay spreads may be long [4]. In other words it is necessary to update the channel estimate in order to improve the performance of the DSRC system. This can be achieved by using data decision feedback to estimate the channel at subcarriers. In additionally, the channel tracking method can be performed in various ways widely investigated in literature but we will examine one of them that called comb co-pilot interpolation.

The paper is organized as follows. The system structure and channel model are presented in section II, and some conventional channel estimation techniques will be given in section III. In Section IV, the EM-SBL Algorithm will be examined and proposed algorithm will be presented. Computer simulations will be carried out in Section V. Finally, contribution and conclusions remarks will be given in section VI.


In the following section, we will briefly outline the scenarios and the IEEE 802.11p structure investigated in this paper. In Figure 1, 802.11p PHY simulation model is shown. In the transmitter part, the short and long training symbols are generated and transmitted at first. Then arbitrary data bits are randomly generated and encoded by convolutional encoder with 1/2 code rate. Then the encoded bits are punctured to support various date rate. The interleaver is exploited to change burst errors into random errors. The interleaved bits are modulated by BPSK, QPSK, 16QAM or 64QAM. The 48 modulated symbols, the 4 pilot symbols (inserted into the subcarriers of -21, -7, 7 and 21), and the 12 null symbols are inserted into 64-point inverse fast Fourier transform (IFFT). After that, 16 cyclic prefix (CP) samples are added to the 64-point output symbol. The CP samples are guard interval (GI) samples to prevent inter symbol interference (ISI) and inter-carrier interference (ICI).

Figure 1. 802.11p PHY simulation model

In the receiver part, assuming perfect time and frequency synchronization, the GI samples are removed before the FFT operation. In order to compensate the channel distortion, equalizer is used for the received data symbols. The demodulation is performed for the equalized data symbols to produce the demodulated data bits, followed by de interleaving and depuncturing operations. Viterbi algorithm using soft or hard decision decoding is used to produce the sink bits which shall be compared with the transmitted bits in order to calculate the BER and PER.

In literature various channel models suitable for vehicular environment have been proposed, some of them are based on extensive measurements taken in outdoor environment. ETSI channel models are an example of propagation scenarios with NLOS component the sparse channel models [5]. Therefore, we choose HIPERLAN/2 model C which is an open-space channel model to simulate the EM-SBL and proposed algorithm performance in a typical urban.


(10 MHz )



NSD: Number of data subcarriers


NSP: Number of pilot subcarriers

52 (NSD + NSP)

NST: Number of used subcarriers, total

0.15625 MHz

ΔF: Subcarrier frequency spacing

6.4 μs (1/ΔF)


32 μs

TPREAMBLE: PLCP preamble duration

1.6 μs (TFFT/4)

TGI: GI (Guard Interval) duration

3.2 μs (TFFT/2)

TGI2: Training symbol GI duration

8 μs (TFFT+ TGI)

TSYM: Symbol interval

16 μs

TSHORT: Short training sequence duration

16 μs

TLONG: Short training sequence duration

Table 1. 802.11p System Parameters


Popular OFDM based wireless applications foresee the transmission of a "preamble" composed of proper training sequences for channel estimation purpose. This is also the case of IEEE 802.11p [6], that specifies the presence of the initial short training symbols (STS) and long training symbols (LTS) inside the PLCP preamble (shown that Figure 2) and of the pilot symbols inside the DATA symbols. These four pilot subcarriers are unable to adequately sample the dynamic channel variation in the V2V setting.

The drawback of LTS method is that the channel will change considerably over the duration of a packet, thus making the initial channel estimate obsolete. Therefore, an appropriate method for tracking the channel and updating the channel estimate is required. For instance, data decision feedback algorithm makes use of the received data to update the channel estimation available at the receiver. The main idea behind data decision feedback channel estimation is the reconstruction of the transmitted OFDM symbol in order to consider the received data as a training sequence and perform channel estimation exploiting the new available information.

In the following part, EM-SBL based sparse channel estimation will be introduced. The conventional estimation algorithm least squares and decision feedback estimator as called comb copilot interpolation will be mentioned while proposed algorithm is explaining.

A. Sparse Channel Estimation with EM-SBL

The EM algorithm is employed, with the transmit data X and as the unknown variables, and the sparse channel vector h as the nuisance variable. The E-step provides the posterior density of the nuisance variable, and hence, the MAP estimate of the sparse channel vector is obtained. Expectation and Maximization (EM) and Sparse Bayesian Learning (SBL) algorithms are combined [9] to propose a joint maximization of over the unknown data symbols as well as the unknown hyperparameters in the M-step.

According to [9] the parametric form of SBL prior can be written as


where constitutes the hyperparameters, which control the variance of each of the channel coefficients and L is length of the complex channel impulse response (CIR). The E and M-steps of the EM-SBL algorithm can be given as,



Figure 2. The packet structure of IEEE 802.11p

B. Proposed Algorithm

The iterative decision feedback equalizer considered here is mainly based on the algorithm from [7] with some modifications. In this section, the EM-SBL algorithm based proposed decision directed method is described. Figure 3 shows the general block diagram for the decision feedback estimator. The proposed algorithm is summarized as follows:

(1) Initialization. The EM-SBL algorithm requires a reasonably good initial estimate of the unknown parameters and to achieve the global maximum instead of a local maximum. Initial estimate for and given by


In conventional method, the initialization of X should be obtained from the channel estimate of the SBL algorithm after p iteration (denoted), but we use copilot channel estimates that will be defined in below instead of.

In the comb copilot interpolation scheme [8], several "copilot" subcarriers are formed from data subcarrier information. Let denote the channel frequency response (CFR) for the OFDM symbol at t. Before forming copilots, this symbol must be equalized with the previous channel estimate, giving


Once this is done, then the channel estimate at a subcarrier k may be formed from the bit decision at this subcarrier, or


where is the decided symbol value generating by soft decision. The copilot channel estimate at subcarrier λ is formed as


where represent weights given to each subcarrier estimate, and . It should be noted that nonexistent subcarriers are excluded from this averaging operation. Additionally, the value of is replaced with an average of the subcarrier and as there is no data transmitted on the zero subcarrier.


It should be note that is the initial channel estimate given by least square method that using long training symbols. The packet structure of 802.11p is shown in Figure 2.

For Least Squares (LS) channel estimation, first and are extracted. Then their N-point DFT are computed as follows.




The LS estimate for is given as follows.


(2) Data detection and CFR update. The EM-SBL algorithm is processed when initialization has been completed. The posterior density computed in the E-step is , where


Note that, using the posterior density is the MAP estimate of the sparse channel vector at the end of the EM iterations. In (12), F is the NxL DFT matrix, X is an NxN diagonal matrix consisting of the N transmits symbols and N is the used subcarrier number. The M-step can be simplified to obtain [9],


where and are the i-th diagonal component and i-th component of andrespectively.


where Y is a NxN diagonal matrix with the received data along the diagonal, and and is the positive square-root of the real-valued quantity


The decision on the data symbols are obtained by mapping the soft decisions to closest point on the constellation,



A 1/2 coded BPSK modulated OFDM based 802.11p system operating at a channel bandwidth of 10MHz is simulated to verify the performance of the combined algorithm. Each packet consists of 13 OFDM data symbols and each OFDM symbol consist 52 used subcarriers (48 data + 4 pilots).

The wireless multipath channel is assumed to follow HIPERLAN/2 ETSI C (L=11 and non-zero taps K = 9). We should note that the tap locations for ETSI C don't match sampling grid of our discrete-time simulation, which is given by the sampling period. So, the tap locations are adjusted the value of the nearest point on the sampling grid of discrete simulation. For EM-SBL algorithm, the initial estimate of X is computed using (19) and the stopping criterion has been chosen as or 25 iterations. The algorithm is compared with training symbol estimation and comb copilot interpolation.


Figure 4. Comparison of algorithm performance stationary multipath channel


Figure 5. 60 km/h HIPERLAN/2 model C. BER performance over time-variant multipath channel

C:\Users\KaNi\Desktop\compare 4.bmp

Figure 6. The BER performance dependence on the packet length


Figure 4 show that LS estimator is an optimum compromise optimum estimator for OFDM systems but, when signal was propagating over time-variant channel, estimator should be design to track the channel variations. Therefore implementation of more sophisticated estimation techniques seems to be an important subject. In Figure 5, it is shown that the proposed or combined EM-SBL algorithm provides a gain and it is more successful about data detection than copilot algorithm over time variant channel. In addition to that gain, proposed algorithm is not affected by the length of the packet.

Combining these two algorithm provide better channel estimation performance. Further, achieving the channel estimation and data detection at the same time, computational complexity is also reduced. In this work, we use a value of so that each copilot is a weighted average of 3 channel measurements and we use the weights and