Priority Based Transmission Rate Control with Neural Network Controller in WMSNs

Wireless Multimedia Sensor Networks (WMSNs) are networks of wirelessly interconnected sensor nodes equipped with multimedia devices, such as cameras and microphones. Thus a WMSN will have the capability to transmit multimedia data, such as video and audio streams, still images, and scalar data from the environment. Most applications of WMSNs require the delivery of multimedia information with a certain level of Quality of Service (QoS). This is a challenging task because multimedia applications typically produce huge volumes of data requiring high transmission rates and extensive processing; the high data transmission rate of WMSNs usually leads to congestion, which in turn reduces the Quality of Service (QoS) of multimedia applications. To address this challenge, This paper proposes the Neural Control Exponential Weight of Priority Based Rate Control (NEWPBRC) algorithm for adjusting the node transmission rate and facilitate the problem of congestion occur in WMSNs. The proposed algorithm combines Neural Network Controller (NC) with the Exponential Weight of Priority Based Rate Control (EWPBRC) algorithms. The NC controller can calculate the appropriate weight parameter λ in the Exponential Weight (EW) algorithm for estimating the output transmission rate of the sink node, and then ,on the basis of the priority of each child node , an appropriate transmission rate is assigned . The proposed algorithm can support four different traffic classes namely, Real Time traffic class (RT class); High priority, Non Real-Time traffic class (NRT1 class); Medium priority, Non Real-Time traffic class (NRT2 class); and Low priority, Non Real-Time traffic class (NRT3 class). Simulation result shows that the proposed algorithm can effectively reduce congestion and enhance the transmission rate. Furthermore, the proposed algorithm can enhance Quality of Service (QoS) by achieve better throughput, and reduced the transmission delay and loss probability.


INTRODUCTION
Most of the research before in the Wireless Sensor Network (WSN) is concerned with scalar sensor networks that measure physical phenomena, such as pressure, temperature, humidity, or location of objects that can be conveyed through low-bandwidth and delay-tolerant data streams.Recently, the focus is shifting toward research aimed to enable delivery of multimedia content, such as audio and video streams, as well as scalar data.This effort resulted in distributed networked systems, referred as Wireless Multimedia Sensor Networks (WMSNs), Akyildiz, et al.,2007.WMSNs are set of sensor nodes, whereby the nodes are equipped with multimedia devices such as cameras, and microphones.Thus a WMSN will have the capability to transmit multimedia data, such as still pictures, stream video, voice, animal sounds, and monitoring data.One important requirement of applications in WMSNs is low delay bounds.Furthermore, some applications of WMSNs need relative resilience to losses.WMSNs can support different types of traffic classes, Akyildiz, et al.,2007.
There are many different resource constraints in WMSNs involving energy, bandwidth, memory, buffer size and processing capability .Given the physically small nature of the sensors, and that multimedia applications typically produce huge volumes of data requiring high transmission rates and extensive processing, this may cause congestion in the sensor nodes.Thus, developing protocols, algorithms and architectures to maximize the network lifetime while satisfying the quality of service (QoS) requirements of the applications represents a critical problem.In most WSN and WMSN applications, traffic mainly flows from a large number of sensor nodes to a base station (sink) node, Akyildiz, et al., 2002.Congestion control is an important issue in transport protocols.Congestion is also a difficult problem in wireless sensor networks.It not only wastes the scarce energy due to a large number of retransmissions and packet drops, but also hampers the event detection reliability.Congestion in WSNs and WMSNs has a direct impact on energy efficiency and application, QoS Ee , and Bajcsy, 2004.Usually, congestion occurs in the bottleneck since it receives more data than it is capable of sending out.In this situation, packets will be queued and sometimes get dropped.As a consequence, response time will increase which causes throughput to be degraded, Samiullah, and Karim, 2011.Two types of congestion could occur in sensor network, as show in Fig. 1 and Fig. 2. The first type is Node -Level congestion that is common in conventional network.it is caused by buffer overflow in the node and can result in packet loss, and increased queuing delay, Malar, 2010.Not only can packet loss degrade reliability and application QoS, but it can also waste the limited node energy and degrade link utilization.In each sensor node, when the packet-arrival rate exceeds the packet service rate, buffer overflow may occur.This is more likely to occur at sensor nodes close to the sink, as they usually carry more combined upstream traffic, Yaghmaee, and Adjeroh, 2008.The second type is Link-Level congestion, which occurs in wireless transmission and occurs when the nodes are in the same utilization channel, for example, Carrier Sense Multiple Access with Collision Detection (CSMA/CD).Such a situation occurs when multiple active nodes perform access on the same channel and collision is then the result, Wan, and Siphon, 2005.

RELATED WORK
Various congestion control techniques have been studied for wireless multimedia sensor networks .thecongestion control mechanisms all have the same basic objective: they all try to detect congestion, notify the other nodes of the congestion status, and reduce the congestion and/or its impact using rate adjustment algorithms. is used to estimate the output transmission rate of the sink node.The FLC is associated with the Exponential Weight (EW) algorithm for selecting the appropriate weight parameter, and then, on the basis of the priority of each child node, an appropriate transmission rate is assigned.Pawarl, and Kasliwal, 2012, proposed a QoS-based sensory Media Access Control (MAC) protocol, which does not only adapts to application oriented QoS, but also attempts to conserve energy without violating QoS-constraints.proposed MAC layer protocol for WMSNs satisfy feature like Maximize network throughput, Enhance transmission reliability, and Minimize control overhead, be energy-efficient, Guarantee a certain level of QoS.
3. THE NETWORK MODEL Fig. 3 shows a simplified experimental model for WMSN .This network model consist of ten nodes, nine sensor nodes, one sink node and the Base Station (BS).The locations of sensor nodes and the base station are fixed.Each sensor node has the knowledge of its own geographic location and the locations of its 1-hop neighbor nodes.Each of the nodes can sense different types of data at the same time and sends those to BS.For each node in the network, there is a single path to reach to the BS.In this network model node may generate different types of traffic .forexample node 9 produces only NRT3 traffic, node 6 produces two type of traffic NRT1 and NRT2, while node 8 produces four type of traffic RT, NRT1, NRT2, and NRT3.The queuing model of each node is shown in Fig. 4.Each traffic class well be buffered in a separate queue, To discriminate traffic classes from each other, the wireless node adds a traffic class identifier to its local sensor packets; hence, when a data packet enters a transmission traffic classifier, the data type will be classified to enter the respective queue that it belongs to, then priority queue scheduler has been provisioned to schedule the diverse traffic with different priority from the priority queues.as shown in Fig. 5, where denotes as the rate of the output of the sink node.And is the sum of input transmission rates for all the childe nodes for transmission data to the sink node.Furthermore, the transmission rate adjusts according to the priority of each child node.In the proposed algorithm, The NC controller follow to estimate the output transmission rate of the sink node, where is the transmission rate for the sink node to transmit data to the BS.In addition, the transmission rate of the node is adjusted according to the priority of the data type and the geographical location of the node .The structure of the proposed algorithm that contains NC controller is based on neural network which is shown in Fig. 6 .The Feed Forward neural network is constructed in three layers.One unit in the input layer, four units in hidden layer, and one unit in the output layers .theoutput signal from the controller is the weight parameter (λ) that is used in EWPBRC algorithm.The activation function of the hidden layer is sigmoid.In Fig. 6 ,W denotes the connection weights between the input layer and the hidden layer , and V denotes the connection weights between the hidden layer and the output layer .The NC controller is trained off-line using The Back Propagation (BP) training algorithm .The simulation of BP algorithm is done using MATLAB program .During the Training process weights in NC controller are adapted to optimize the controller response.The congestion control unit in the proposed NEWPBRC algorithm is shown in Fig. 7, when input rate passes through the CDU unit, calculate error, and then after the adjustment of the output transmission rate by the NEWPBRC controller and RMU unit, a new rate is generated to adjust the rate for transmitting from the sink node to the BS and the transmission rate for transmitting from all the child nodes to the sink node.

. RATE MANAGEMENT ALGORITHM
NEWPBRC algorithm used for adjustment the transmission rate while congestion occurs.This algorithm can be divided into three steps: Step 1: Computing the new output transmission rate of sink node.
achieved by the proposed algorithm is low by (13%) than that achieved by FEWPBRC algorithm.The proposed algorithm can be (4.61%)less average loss probability than FEWPBRC algorithm.

Sink node all child node
Step 3 : Computing a new output transmission rate for each parent node Step 1 : Computing the new output transmission rate of sink node .
Step 2: Computing the new output transmission rate of child node .

Priority Based Transmission Rate Control with Neural Network Controller in WMSNs
Nadia Adnan Shiltagh Ali H. Wheeb Fig. 5 shows the block diagram of the proposed algorithm which is called Neural controller Exponential Weight Priority-Based Rate Control (NEWPBRC), for the proposed algorithm which combines the Neural Controller (NC) with the Exponential Weight of Priority-Based Rate Control (EWPBRC) algorithm.The NC controller is used to adjust the weight parameter λ in the EWPBRC algorithm to obtain the optimalas shown in Fig.5, where denotes as the rate of the output of the sink node.And is the sum of input transmission rates for all the childe nodes for transmission data to the sink node.Furthermore, the transmission rate adjusts according to the priority of each child node.In the proposed algorithm, The NC controller follow to estimate the output transmission rate of the sink node, where is the transmission rate for the sink node to transmit data to the BS.In addition, the transmission rate of the node is adjusted according to the priority of the data type and the geographical location of the node .The structure of the proposed algorithm that contains NC controller is based on neural network which is shown in Fig.6.The Feed Forward neural network is constructed in three layers.One unit in the input layer, four units in hidden layer, and one unit in the output layers .theoutput signal from the controller is the weight parameter (λ) that is used in EWPBRC algorithm.The activation function of the hidden layer is sigmoid.In Fig.6,W denotes the connection weights between the input layer and the hidden layer , and V denotes the connection weights between the hidden layer and the output layer .The NC controller is trained off-line using The Back Propagation (BP) training algorithm .The simulation of BP algorithm is done using MATLAB program .During the Training process weights in NC controller are adapted to optimize the controller response.The congestion control unit in the proposed NEWPBRC algorithm is shown in Fig.7, when input rate passes through the CDU unit, calculate error, and then after the adjustment of the output transmission rate by the NEWPBRC controller and RMU unit, a new rate is generated to adjust the rate for transmitting from the sink node to the BS and the transmission rate for transmitting from all the child nodes to the sink node.

Yaghmaee, and Adjeroh, 2009, proposed
Priority Based Rate Control Algorithm (PBRC) used for congestion control and service differentiation in WMSNs.

Chen, and Lai, 2012, proposed
an algorithm where a Fuzzy Logical Controller (FLC)