Books - Non-fiction

Implementation of Hybrid ARQ (HARQ) Error Control Algorithm for Lifetime Maximization and Low Overhead CDMA Wireless Sensor Network (WSN)

Description
The deployment of high densities of node due to the advancement in Wireless Sensor Network (WSN) technology had created the concern regarding the lifetime and error presented in the network. From the previous studies, Hybrid ARQ (HARQ) error control
Published
of 6
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
Share
Transcript
  Implementation of Hybrid ARQ (HARQ) Error Control Algorithm for Lifetime Maximization and Low Overhead CDMA Wireless Sensor Network (WSN) Samirah Mohd Razali, Kamaruddin Mamat, Nor Shahniza Kamal Bashah Faculty of Computer and Mathematical Sciences Universiti Teknologi MARA 40450 Shah Alam, Selangor, Malaysia  Abstract   — The deployment of high densities of node due to the advancement in Wireless Sensor Network (WSN) technology had created the concern regarding the lifetime and error presented in the network. From the previous studies, Hybrid ARQ (HARQ) error control techniques can combat errors and indirectly, can reduce the energy consumption of sensor nodes. Based on the latest previous work, the HARQ algorithm was not able to combat the issue of interference levels properly that rises to Coded Division Multiple Access (CDMA). Not to mention, the implementation of high error correcting capabilities of Error Correcting Codes (ECC) in the module of HARQ on the unconducive environment will cause additional overhead. Thus, the problem regarding the interference and overhead in WSN becomes our motivation to research a better HARQ error control algorithm that follows the process of convectional Automatic Repeat Request (ARQ) and Forward Error Correction (FEC) with network condition estimation in CDMA WSN. This paper evaluated the different type of error correction schemes and existing HARQ over different Bit Length and Node Densities. This paper had demonstrated that as the network get congested, the error rates will get increased. The congested network had also increased the energy consumption that will shorter the lifetime of the network.  Keywords-wireless sensor network; hybrid-ARQ; FEC; Coded  Division Multiple Access; Energy I.   I  NTRODUCTION  The critical benefits of Wireless Sensor Network (WSN) as scalable, cost-effective technology had turned many contributors, researchers and even vendors in implementing WSN in applications such as monitoring, tracking, medical, agricultural and many more. WSN crucial in monitoring field to monitor natural disasters such as flood, volcano eruption, earthquake and not to mention, very critical to track water current of the river for early flood detection and warning system. The previous implementation of such application had  been studied by many researchers such that in [1] and [2]. As the technology evolved from years to years, more nodes are deployed from hundreds to even thousands of sensors and more heavy data are transferred through WSN. The existing architecture of WSN cannot tolerate the advancement in which there will cause high error rates when network starts to get congested. The CDMA architecture was implemented instead of existing Media Access Control (MAC) as CDMA might be able to optimize the overall performance of WSN in taking concern of delay, fault-tolerance, that MAC does not  provide which also being stated in the work of [3]. The previous study of [4] [5] [6] show that the HARQ is one of the protocols that exploits the benefit of both ARQ and FEC to lower the Bit Error rates (BER) and maintain the energy consumption for CDMA WSN. However, the paper had not resolved the problem of interference effect that rises due to CDMA Multiple Access (MAI) and Node Interference (NI). Besides, end-to-end error correction fosters higher BER in low hop count. As for this problem, the new approach in which designed the new HARQ algorithm with network estimation to control transmit power and proper error correction schemes at receiver were proposed in the paper of [7]. This paper evaluates the performance of CDMA WSN network that implement the existing process of HARQ, in which cooperating end-to-end error detection and correction. We cooperated the existing CRC-4 with various type of ECC to proof and study the effectiveness of different ECC on different condition of the network. We also test the multiple ECC implementation that follows the paper of [7] in terms of  Energy consumption and  BER. Based on [7], the error correction schemes will be changed according to the change in network condition. The network condition will be estimated using Kalman Filter and the estimation value is in the form of SNR will be used to decide the error correction schemes that will be initiated by the sender as well as error correcting capabilities. However, in this paper, we do not cooperate the estimation technique and transmit power that was proposed in [7], but we only test the multiple ECC mechanism that was demonstrated for Scenario 4 and explained in details in Method Section. The evaluation of the experiment conducted will be made on different  Bits Length  and  Number of Nodes . The rest of this paper is organized as follows: the related work is explained in Section II. The method of implementation and the experiment for existing schemes that had been carried out were stated in Section III. In Section IV, the results for implementation of existing schemes were  presented and analysed. The discussion is carried out in Section V. In Section VI, conclusions and future work are given out. II.   R  ELATED WORKS  This research works follow the paper by [7] [8] in order to implement the new algorithm of HARQ. The paper related to the study of HARQ such as [5]. In this paper, the author had implement HARQ with BCH in multi-hop communication. 2016 IEEE Conference on Wireless Sensors (ICWiSe)978-1-5090-1626-6/16/$31.00 ©2016 IEEE 71  The paper by [9] [10] [11] uses the Transmission Power Control (TPC) as the method to extend the lifetime of the WSN. The idea to reduce interference effect in CDMA using transmit power control had been stated by [12] [13]. There is also a paper by [8] that had demonstrated that the usage  between error control protocols HARQ with TPC can be significant to improve the network of sensors. However, in their paper, the effect of combining both HARQ and TPC does not have much significant effect on the energy consumption as HARQ and TPC denoted different algorithms and techniques and combining these two techniques without a  proper study might give some cancellation effects to the network. However, transmit power can reduce or combat the effect of interferences [14] [15] such as MAI [13] and NI that rises due to CDMA. Thus, incorporating transmit power in the  process of HARQ is still consider promising technique to lower overhead rather than incorporating separated algorithm of TPC at the sender or receiver. Based on the study of Masood, the authors had demonstrated the estimation using Kalman Filter on the channel condition to control the transmission power to reduce  power consumption effectively. The power will be adjusted according to environment condition. Kalman Filter is one of the critical Bayesian filtering [16] [17] that besides, it minimizes a quadratic function of estimation error for a linear dynamic system with white measurement and disturbance noise. [18] III.   M ETHODS  This research simulated four scenarios using MATLAB which had been stated in Table 1. The scenarios as follows; Scenario 1: Without any error detection or correction, Scenario 2: With various Error Correcting Codes implemented but without retransmission, Scenario 3: Retransmission only implemented and Scenario 4: with retransmission and various error correcting codes implemented. The error detection was using CRC-4 and Error Correcting Codes such that BCH, Reed-Solomon (RS), and Convolutional Codes were tested for the scenarios that implemented ECC. The input data were generated in random  binary numbers. For the transmission that was coded with ECC, the input data will be passing through ECC encoder such as RS encoder, BCH encoder and convolutional encoder. Matlab built-in CRC Generator were implemented using CRC-4 and BPSK modulation were taking place. Next, spreading were carried out for the assignment of Pseudo Noise (PN) code for each user defined in the network. For the transmission that does not implement the ECC, only the CRC-4 will be appended to the input binary data before modulated, spread and passed through the channel. The CRC decoder will calculate and display the bits that have errors while RS decoder and BCH decoder will decode and correct errors that occurred in the receiving data. The data that were encoded with convolutional codes will be decoded with Viterbi decoder. Table 1 indicates the defined variables and their values implemented in this research. We test all the variables and  parameters against two metrics (BER and Energy Consumption) in order to study the effect of the different error correction schemes, different bit length, and different error correcting capabilities. These aspects have significant effect on the network such that overhead rises due to error correcting codes, interference present, and architecture of the network. These aspects need to be seriously taking into account in designing new algorithm and cannot be cast aside. We have conducted the simulation using Matlab R2015b under Window 10 64-bit. The simulations consist of a few experiments that follow all corresponding scenarios defined. The value of node densities and packet length were changed during the experiments following Table 1 below. The Scenario 1 consists of one experiment which the network simulate does not implement any error detection or correction. Scenario 2 consists of three experiments which follow the different error correcting codes such as RS, BCH and Convolutional codes respectively. Scenario 3 consists of one experiment which simulates only the retransmission once the errors were detected and count the number of retransmission need to be initiated to complete a successful transmission without corrupted bits. And, lastly, the scenario 4 consists of three experiments that implement different ECC with retransmission had set the method of error correcting as a survey following our proposed algorithm that was defined in [7]. However, we do not cooperate any estimation techniques or transmit power that was defined in [7]. We classified the network condition into three conditions such as less congested, optimal condition and highly congested which depending on the SNR value. We assume as the SNR get higher, the network condition is less congested. The initial transmission was initiated to count the number of bit errors that occurred during the first transmission from sender to receiver. The multiple error correction scheme will  be responded to the network condition and the number of errors detected at the first transmission. At congested network (SNR is lower than 14 dB), the receiver is set to perform error correction using BCH and RS. Otherwise, the retransmission will occur. For example, at higher SNR such as 14db to 18db, the sender and receiver are set to perform retransmission if errors occurred. While, at 8dB to 12dBin which we consider as the optimal condition for a network, we set RS (15,3) for the error correction and at lower SNR of 2dB to 6dB, we set BCH (15,5) to correct errors. For the second transmission, we consider the SNR values and also the number of errors that were detected during the first transmission. TABLE I. T ABLE OF REMAINING ENERGY FOR ALL ERROR CORRECTION SCHEMES AND NO .  OF NODES   Variables Values Channel Model Rayleigh Fading CDMA Multi-Carrier CDMA (MC-CDMA) Modulation BPSK  Noise AWGN Error Detection CRC-4 Error Correction Codes Reed-Solomon, BCH, Convolutional Codes  Number of Nodes 4,16,32,48,64,80 Bit Length (Bits) 10000,15000,20000,25000,30000 Minimum Distance bet. Sender and Receiver (meter) 5 meter Path Loss Exponent 3.5 Payload (Bits) 256 2016 IEEE Conference on Wireless Sensors (ICWiSe) 72  Header (Bits) 128 Area of monitoring field (meter) 500meter x 500meter Scenarios/Error Correction Schemes Scenario 1 : CDMA WSN without any error correction scheme Scenario 2 : CDMA WSN with CRC-4 and Error Correcting Codes (BCH, RS, Convolutional Codes) Scenario 3 : CDMA WSN with retransmission only Scenario 4 : CDMA WSN with retransmission and with Error Correcting Codes (BCH, RS, Convolutional Codes) We formulate the BER calculation, Received Signal and Energy Consumption. The expression of BER or probability of error,    for BPSK is by using the standard equation as shown in equation (1) below [19]: =   (       (1) Thus, the expression of BER for BPSK in Rayleigh Fading denoted as shown in equation (2) below [19][20]: BER    =   (1−    )   (2) While, received signal power follow free space path loss cooperating the path loss exponent and the distance between transmitter and receiver. Given the received signal, Y   in Additive White Gaussian Noise (AWGN) as in equation (3) [21] [22] below and taking Rayleigh Fading as consideration as in equation (4): =ℎ()+  (3) Where ℎ  is channel gain,   is transmitted data and   is noise. Channel gain, ℎ  captures the effect of Rayleigh fading [23] in which the channel gain, ℎ  can be formulated again as in equation (4): Y=h (d  )  (4) Where  , the path loss exponent is denoted that 2≤ ≤6  while   is defined as the distance between transmitter and receiver (in meters). The expression of minimum energy, E are as shown in equation (5) in which each transmitted bit consume 1 unit of energy and received bit consume 0.75 units of energy according to [24]. =      (  +     0.75)  (5) Where H is the number of hops,    is the number of packets, and    is the total number of bits in which include the header and payload. The minimum energy also denoted the transmission without any error detection or correction applied. Following the equation from [24], the energy consumed in the transmission with retransmission as in equation (6):   =      (  +   + (   +  )  0.75+  )  (6) Where    is the decoding energy. The encoding energy are so small that it can be negligible. [25] However, this paper only implements the end-to-end error detection and correction which add CRC-4 for error detection and no ACK or NACK message were applied to the transmission. Thus, the equation for energy consumed in a transmission for CRC-4 and Error correcting codes can be rewritten as below:   =      (  + (    0.75)+    (7) Based on [26] [27], the    can be expressed as:   =(2+2  )(  +   )  (8) Where n is block length for corresponding error correcting codes and t is error correcting capability. While,   +   correspondence to the energy consumed in addition and multiplication of error correcting codes. IV.   R  ESULT AND FINDING  The results of the implementation of existing error detection and correction are to analyse the effects of different error correction schemes on the number of nodes and bit length towards BER and energy consumption. This collected results can be a benchmark to initiate the proper modification of HARQ algorithm to follow the approach in the paper of [7]. Fig. 1 below compares the average BER for all error correction schemes with CRC-4 with BCH, RS and Convolutional Codes and without error correction codes. The correcting codes used in this experiment have medium error correcting capabilities, such that for BCH (15, 5), error correcting capability, t=3. While, for RS (15, 5), error correcting capability, t=5 and for convolutional code rate is ½. The graph had demonstrated that the medium error correcting capability of BCH able to correct the error when SNR approximately reach 16, such that the BER had fallen to 0.3084. While, for the other error correction schemes such as RS and convolutional codes, higher error correcting capabilities need to be considered to correct errors. Figure 1. Graph of average BER all error correction and without any error correction for increasing SNR. TABLE II. T ABLE OF REMAINING ENERGY FOR ALL ERROR CORRECTION SCHEMES AND NO .  OF NODES   No. of  Nodes (Users) Remaining Energy without Any Error Remaining Energy with CRC-Remaining Energy with CRC-Remaining Energy with CRC-4 and Convolution 2016 IEEE Conference on Wireless Sensors (ICWiSe) 73  Correction Codes (μA) 4 and BCH codes (μA) 4 and RS codes (μA) al codes (μA) 4 23053.0 21420.0 21239.0 23032.0 16 2881.6 5355.1 5309.8 5758.0 32 2881.6 2677.6 2654.9 2879.0 48 1921.1 1785.0 1769.9 1919.3 64 1440.8 1338.8 1327.4 1439.5 80 1152.6 1071.0 1062.0 1151.6 Table 2 above shows the overall values of remaining energy after transmission took place from sender to receiver. The initial energy configured is at 100000 μA in which equal to 100 mA. We consider a very low calibration as for the data were sent in Bits. Figure 2. Graph of Remaining Energy against increasing number of nodes for different error correction schemes The analysis of remaining energy was shown in Fig. 2.   The remaining energy for the network without intervention of error correction schemes shows the lowest possible usage of energy for the lowest number of nodes and increase slightly as the number of nodes also increased. The lowest possible energy usage recorded as RS code was being implemented.   However, considering the BER in Fig. 3, RS code could not handle the errors as compared to BCH. Figure 3. Graph of average BER against SNR for increasing length of Bits Fig. 3 depicted the average BER for increasing number of Bits from 10000 bits to 30000 bits. The graph had shown that as the number of bits in transmission increased, the BER will also increase. Not to mention, the intervention of CRC error detection that appends the extra bits to the transmitted data also have chances to increase the BER in the network. The modulation scheme or PN code generation that is a part of CDMA architecture to assign the PN code to users might add additional bits to the network which might lead to the increase in BER. Thus, the architecture such that modulation and spreading need to be considered properly to minimize the  potential overhead and BER. Figure 4. Graph of average BER for different error correction schemes with increasing Bit Length. Fig. 4 shows the Average BER for different error correction schemes with increasing Bit Length at SNR=20. The lowest BER was obtained using BCH codes. The convolutional codes induce more BER in this case due to the addition of unnecessary redundancy and the inability of convolutional codes to handle the errors. However, at the constant SNR=20, the BER remain almost constant with the increasing number of Bits. This is might due to the difference  between the increased bits that were so small which do not add significant effects on the network BER. Based on the graph in Fig. 5, the remaining energy decreased as the number of bits increased. The error correction schemes that had the possible lowest energy usage are convolutional codes and BCH while the RS used the most energy. However, the convolutional codes might use lower energy because it did not handle any errors as compared to other error correction schemes as shown in Fig. 4 where BER of network using convolutional codes were the highest. Figure 5. Graph of remaining energy for different error correction schemes with increasing Bit Length. 2016 IEEE Conference on Wireless Sensors (ICWiSe) 74    The retransmission initiated once the errors were detected using CRC detector with the increasing number of nodes and  packet length as shown in both Fig. 6. Figure 6. Graph of average number of retransmission vs. number of nodes with a constant number of bits = 10000. Based on Fig. 6, the increase in the number of nodes have increased the number of retransmissions. This is because, as more nodes being added to the network, the network will get congested and causes the increment of BER. Thus, more than a single retransmission needed to be initiated for a successful delivery of the whole transmitted bits. As well as in the condition where there is the increase in the number of bits with a constant number of nodes, the higher number of bit sent in the network also promoting to the increment of the average number of retransmission throughout the experiment. Fig. 7 below shows the graph of average BER after the first transmission for transmission that implement BCH (15,5) and also transmission that implement multiple Error Correction. The scheme using BCH (15,5) shows a better degradation instead of Multiple ECC. Figure 7. Graph of average BER for BCH (15,5) and Multiple ECC against increasing SNR. (Scenario 4) However, as shown in Fig. 8 below, the used of BCH (15,5) throughout all the network condition were not so  promising towards the energy consumption issues. The multiple ECC scheme shows higher remaining energy at higher SNR as higher SNR uses only retransmissions. Figure 8. Graph of remaining energy for BCH (15,5) and Multiple ECC against increasing SNR. (Scenario 4) V.   DISCUSSION From the result obtained, we had proved that the BCH (15, 5) codes can tolerate the errors in the improved network condition where there less possibilities of congestion. The high BER at the lowest SNR had shown that the error correcting codes of BCH (15, 5), RS (15, 5) and ½ convolutional codes are unable to correct errors as the network without the use of error correction schemes had a significantly lowest BER. The BER increased when these error corrections are applied due to the redundancy added to the network. For the BCH (15, 5), the redundancy added to the network approximately around 20000 bits. Even though BCH able to correct the BER as SNR improved, but the condition when the network congested need to be considered. The using of much higher error correcting capability such that t=7 or t=10, might add more redundancy to the network. The depletion of energy can be caused by the addition of redundancy due to the high error correcting capabilities as well as congestion due to the increasing number of nodes. This is because there is more energy needed to transfer a certain number of bits with the addition of redundancy that append to the transmitted data. The decoding process used more energy than the encoding process which causes most of the high energy usage. From the result obtained, the higher error correcting capability is not the only aspect that causes the high energy depletion but also the aspects of code rate and design. The RS (15, 5) have t = 5 while BCH (15, 5) have t=3 such that the energy consumed by BCH a slightly lower than RS even the gap is so small. For the retransmission scheme that was carried out in Scenario 3, the retransmission strategy was much beneficial for the uncongested network by means of lower node densities or smaller packet sizes. The higher number of retransmission were recorded if the network start congested or packet length increases. For scenario 4, the implementation of multiple ECC has not been able to reduce BER as compared to the implementation of BCH (15,5) throughout all network condition. This is due to the lower code correcting capabilities of RS (15, 3) and BCH (15, 5) implemented. However, from the result obtained for scenario 4, the multiple ECC schemes show a promising result toward the energy consumption issues as compared to BCH (15,5). This is 2016 IEEE Conference on Wireless Sensors (ICWiSe) 75
Search
Similar documents
View more...
Tags
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks
SAVE OUR EARTH

We need your sign to support Project to invent "SMART AND CONTROLLABLE REFLECTIVE BALLOONS" to cover the Sun and Save Our Earth.

More details...

Sign Now!

We are very appreciated for your Prompt Action!

x