Pre-channel scheduling and Priority-based reservation in medium access control for industrial wireless sensor network applications

ARTICLE IN PRESS

JID: CAEE

[m3Gsc;May 13, 2017;16:37]

Computers and Electrical Engineering 0 0 0 (2017) 1–15

Contents lists available at ScienceDirect

Computers and Electrical Engineering journal homepage: www.elsevier.com/locate/compeleceng

Pre-channel scheduling and Priority-based reservation in medium access control for industrial wireless sensor network applicationsR Kannan R. a,∗, Solai Manohar S. b, Senthil Kumaran M. c a

Faculty of Electrical Engineering, Anna University, Chennai 600025, India Department of EEE, KCG College of Technology, Chennai, India c Department of EEE, SSN College of Engineering, Chennai, India b

a r t i c l e

i n f o

Article history: Received 8 November 2016 Revised 26 April 2017 Accepted 8 May 2017 Available online xxx Keywords: Data fusion Industrial Wireless Sensor Network (IWSN) Packet classification Reservation-based MAC Slot scheduling

a b s t r a c t Numerous technical advancements and the pervasive controlling schemes in the Industrial Wireless Sensor Network (IWSN) with the capability of interoperability among the nodes facilitate the reliable communication. With the increase of participating sensors size, the memory and energy consumption reduce the lifetime adversely. To alleviate these issues, this paper creates an energy efficient data management and routing architecture based on the data fusion techniques in IEEE 802.15.4. The data classification through proposed MultiStage Classification (MSC) prior to priority assignment techniques makes the immediate decision regarding the controlling actions. The split up of data using Reservation-based Medium Access Control (RMAC) raises the resource allocation requests according to the data type. The built up of the relationship between the number of slots and the data arrival enhances the resource utilization performance for each slot. The guarantee of transfer of immediate transmission with the priority supports the critical conditions with high energy efficiency. © 2017 Elsevier Ltd. All rights reserved.

1. Introduction With the numerous deployment of tiny and intelligence sensor devices, the application of IoT is extended to support the real-time applications drastically. The operational constraints (resource utilization, reliability) are diverse in nature like the controlling of one will affect the other parameter. Hence, the monitoring and maintenance of the sensor devices in a unified manner are the challenging task in Industrial Wireless Sensor Network (IWSN) applications. The important issue to achieve the reliable communication among the sensor nodes is the target coverage problem. Besides, the uses of feedback control loops in the industrial environment highly impose the end-to-end delay during data transmission among the devices in the IWSN. Due to the inheritance of broadcast characteristics, the wireless medium is opened to the either authorized or unauthorized users (eavesdropping attacks). The challenging issue with the real-time reliable communication is to meet latency requirements for the feedback control loop. The multi-hop communication governed by the industry standard called IEEE 802.15.4 increases the communication delay under dynamic challenges in channel conditions.

R ∗

Reviews processed and recommended for publication to the Editor-in-Chief by Associate Editor Dr. M. H. Rehmani. Corresponding author. E-mail address: [email protected] (K. R.).

http://dx.doi.org/10.1016/j.compeleceng.2017.05.010 0045-7906/© 2017 Elsevier Ltd. All rights reserved.

Please cite this article as: K. R. et al., Pre-channel scheduling and Priority-based reservation in medium access control for industrial wireless sensor network applications, Computers and Electrical Engineering (2017), http://dx.doi.org/10.1016/j.compeleceng.2017.05.010

JID: CAEE 2

ARTICLE IN PRESS

[m3Gsc;May 13, 2017;16:37]

K. R. et al. / Computers and Electrical Engineering 000 (2017) 1–15

The reliability enhancement depends on the scheduling of slots, requests, frequency hopping, mesh-star routing and the redundancy routing. Sensor Nodes (SN), Relay Nodes (RN) and the Base Stations (BS) are the major parts of the IWSN. The deployment of RNs on single connected environment ensures the connectivity between the sensor nodes and the decisionmaking the unit. But, the failure in RN under the harshest environments breaks the functionality of the network and disconnects the data forwarding link to the base stations. Resource constraints, rescheduling and the routing ordered slot assigning raises the challenges in slot scheduling. The extension of IWSN into the cognitive models is the active research area nowadays. The design principles of Cognitive Radio Network (CRN) models [1,2] reveal the communication protocols highlight the challenges in node deployment, strategies, clustering issues in detail. The spectrum utilization is the major issue in the CRN formulation that makes the spectrum sensing as the essential mechanism that induce serious threat to sense the spectrum effectively. The security threats [3] observed in the CRN communication were incumbent emulation and the sensing data falsification. The availing of large bandwidth through the channel bonding algorithm [4,5] reduces the contiguous channels effectively. The governing of the Medium Access Control (MAC) layer plays the major role in QoS provisioning in IWSN. But, the lack of priority leads to difficulty in decision making regarding the control parameters. The design of hybrid MAC considers the allocation of sensor node requests in contention periods and the contention free period to meet the diverse demands for different nodes. But, the traffic information in the superframe structure increases the overhead that will deteriorate the performance in IWSN. The standard IEEE 802.15.4 contains two modes such as nonbeacon and beacon enabled modes in MAC. In the first mode, each node sends their data by using the Carrier Sense Multiple Access/Collision Avoidance (CSMA/CA). The clustering-based adaptive routing and data gathering algorithm [6] addresses the issues in transmission and gathers the real-time big data to analyze the risk in the industrial operations. The extension of data gathering is to perform the fault diagnosis with low energy consumption and high network lifetime in industrial applications. Many routing algorithms contribute towards the network lifetime improvement such as Low Energy Adaptive Clustering Hierarchy (LEACH) [7] combines the energy-efficient-based clustering and media access for data aggregation. The self-configuring capability of large number of nodes and algorithms for even distribution of energy support the lifetime improvement. The major disadvantage of LEACH is additional energy overhead due to the transmission of location and residual energy of each node to the decision making unit. Alternatively, the Hybrid Energy-Efficient Distributed (HEED) [8] periodically selects the Cluster Head (CH) on the basis of the residual energy and proximity to the neighbors. Low-message overhead and the preservation of connectivity are the major observations in HEED protocol. The neighborhood informationbased cluster formation caused the uneven energy overheads. The data from the multiple sensors are combined to enhance the accuracy of measurement that initiates the fusion techniques. The detailed study of the protocols implemented on IWSN highlights the challenges as follows: assurance of limited resource usage, satisfying of Quality of Service (QoS) requirements, minimum data redundancy, scalability and security constraints. The major contributions of the proposed work are listed as follows: • The core node selection based on the High-Priority Indication Space (HPIC) super frame mitigates the overlapping in channels and monitors the operations (transmission and reception of data) performed by the nodes. In traditional method [9], the trade-off between the energy efficiency, throughput and the delay is the major limiting factor due to the multidomain constraints of the sensor devices. The introduction of parallel data fusion techniques with the priority assignment scheme provides the trade-off between those parameters. • The employment of Multi-Stage Classification (MSC) before the priority assignment serves as the base for immediate decision making regarding the normal and isolate data. • The data classification (normal and emergency) facilitates the reservation of slots improves the throughput performance with less time and energy consumption. The paper organized as follows: The detailed description about the related works on energy-delay aware routing protocols in IWSN is discussed in Section 2. The implementation process of Multi-Stage Classification (MSC) on Reservation-based MAC layer is described in Section 3. The comparative analysis of the proposed approach with existing delay-aware routing protocols provided in Section 4. Finally, the conclusions about the application of proposed work on various communication scenarios presented in Section 5. 2. Related work 2.1. IWSN scenario A unified management of wireless sensor devices requires the industrial authorities with the capabilities of provisioning the network infrastructure support to IWSN applications. Distributed and dynamic topology of WSN introduced the special requirements for routing protocols. Pantazis et al. [10] provided the survey of energy-efficient routing protocols based on four major schemes such as network model, communication model, topology and reliable routing. The major categories of the protocols are investigated to balance the energy level. Akhtar [11] described the survey of potential renewable energy resources with the characteristics of WSN, battery recharging techniques and the applications of WSN with the challenges and future scope. The incorporation of WSN in an urban environment was investigated in [12]. The limitations of the IWSN applications are energy-conservation during the data gathering by Hybrid Rapid Response Routing (HRRR). With these gathering requirements and the additional features of TinyOS [13] offered the operational flexibility. Please cite this article as: K. R. et al., Pre-channel scheduling and Priority-based reservation in medium access control for industrial wireless sensor network applications, Computers and Electrical Engineering (2017), http://dx.doi.org/10.1016/j.compeleceng.2017.05.010

ARTICLE IN PRESS

JID: CAEE

K. R. et al. / Computers and Electrical Engineering 000 (2017) 1–15

[m3Gsc;May 13, 2017;16:37] 3

2.2. Fault diagnosis quality improvement Hou et al. [14] explored the two-step classifier approach by using the Dempster-Shafter theory to improve the quality of fault diagnosis. The utilization of Dempster-Shafter theory efficiently reduced the payload transmission and energy consumption. Reliability and delay requirements pose the challenges in the dynamic and cluttered industrial environments. The utility of cooperative relays supports the reduction of outages in fading areas. The trade-off between the packet delivery ratio and less selection overhead is necessary. The analysis of test data showed that the study of characteristics of signal transmission was the major requirement. Ding et al. [15] proposed the Real-Time Big Data Gathering (RTBDG) algorithm on the basis of indoor WSN risk analysis. Based on the measure of Received Signal Strength Indicator (RSSI) and the residual information, screening and clustering of the data transmission were established. They also highlighted the demerits of centralized algorithm. The critical demands on reliability and real-time performance raised the several challenges in industrial automation applications. 2.3. Energy conservation issues in IWSN Energy conservation with the strict real-time constraints are the major parameters for the design of IWSN in condition monitoring applications. Zheng et al. [16] proposed the MAC protocol with the multi-user support on the basis of different priority levels. With this protocol implementation, they ensured the gain of channel access by the user having the highest priority. To meet the demand for real-time feedback control, the Wireless Computing System (WCS) requires an acceptable delay for data collection. Low power consumption to reduce the cost was the major objective in the design of IWSN. Suto et al. [17] proposed the energy efficient –delay aware WCS to provide the trade-off between the power. The occurrence of interferences during the packet transmission limits the throughput performance. Diverse QoS requirements such as reliability, delay and throughput for the heterogeneous traffic observed in health care monitoring devices required the adaptive MAC protocols with load constraints. Anjum et al. [18] addressed the above QoS requirements by using the Priority Load Adaptive (PLA) to maintain the efficiency with less power consumption. The simultaneous classification of packets and the calculation of priorities in PLA-MAC protocol provided the necessary scheduling to reduce the power consumption. 2.4. Slot assignment techniques in IWSN The operation of IWSN was affected due to the dust, electromagnetic interferences and the interferences from the other devices. Yang et al. [19] presented improved the reliability in data transmission by using three concepts such as segmented slot assignment, fast slot computation and free node concept. The employment of limited shared slots by using the above three graphs, improved the retransmission efficiency for TDMA-based multihop transmission. Evenness in the spectrum usage was still a major issue in the IWSN-based industrial automation applications. Han et al. [20] analyzed the characteristics of recent energy-efficient strategies and presented the comparisons among them. They showed that the optimal network coverage with some properties under noisy environments to select the appropriate strategy. The analysis of the characteristics of data in real-time monitoring applications (health care system) requires the discrimination of sensor data into either periodic or burst data. They derived closed-loop expressions for the probability of intercept occurrence, event in the roundrobin and optimal sensor scheduling. Besides, they conducted an asymptotic intercept probability analysis to provide the insight view of sensor scheduling. 2.5. Timeslot assignment The occurrences of interference between the neighbor nodes caused the delay in data transmission. The introduction of timeslot assignment and the effective channel utilization in Time Division Multiple Access (TDMA) mitigated the conflicts in packet transmission. Due to the dynamic wireless conditions in the industrial plants, the real-time performance was limited. Lu et al. [21] reviewed the series of recent advances such as scheduling algorithms, new protocols, cyber-physical co-design of simulator under the evaluation of wireless control systems. Node failures during the harsh environment construction require the high installation cost. The detailed literature review presented in this section conveyed that the trade-off between the high throughput and low energy are the major issues in the IWSN-based industrial monitoring. This paper overcomes the issues in traditional protocols with the combination of fusion techniques [22,23] with the priority assigned through the recursive process. 3. Multi-Stage classification and reservation-based MAC This section discusses the implementation of proposed energy efficient and reliable data transmission structure by using the Multi-Stage Classification (MSC) and Reservation based MAC (RMAC) as shown in Fig. 1. Initially, the core node selection among the number of sensor nodes is the first stage of proposed work. The high priority indication space construction on time slices and the estimated boundary values decide the core node in the network. Once the core node is elected, it is responsible for the data forwarding to sink node according to the severity level. The estimation of severity level depends on the data classification and the reservation of slots. Hence, the next stage is the MSC that employs the pre-processing, Please cite this article as: K. R. et al., Pre-channel scheduling and Priority-based reservation in medium access control for industrial wireless sensor network applications, Computers and Electrical Engineering (2017), http://dx.doi.org/10.1016/j.compeleceng.2017.05.010

ARTICLE IN PRESS

JID: CAEE 4

[m3Gsc;May 13, 2017;16:37]

K. R. et al. / Computers and Electrical Engineering 000 (2017) 1–15

Fig. 1. Architecture of proposed work. Table 1 Traffic categories. Priority

Actions needed

1(High) 2 3 4(Low)

Emergency safety action Extremely critical control Critical control Normal data

equivalent class set creation, index estimation and the storage optimization. The sequential processes in MSC split the data into two either normal or emergency. The normal data have undergone the fusion process and stores the normal data into the buffer for future purpose. If the data is an emergency, then the slots allotted in scheduling process require the reservation space. This reservation is based on the three measures such as priority, severity index and volume. Based on these measures, the slots are reserved for an immediate transmission of emergency data to take the necessary control action by the decision unit. The detailed descriptions of major operational stages in the proposed MSC-RMAC are presented in this section. 3.1. Core node selection and scheduling The provision of guaranteed channel access to real-time IWSN applications is done by Time Division Multiple Access (TDMA) slots. The unit in TDMA that allows the transaction once refers time slots. The scheduling algorithm utilizes the time slot series called super frame durations. The allocation of bandwidth to the communication link by using the super frame durations provides the ability of periodic chance to each link. The core node selection among the number of sensor nodes on the basis of priority space in time slots is the initial stage of proposed work. The division of slots into sub-slots according to the minimum time unit (High-Priority Indication Space (HPIS)) serves as the base for emergency data handling. The priority assignment on the basis of actions is defined in Table 1. Fig. 2(a) and (b) illustrates the architecture of time slots and the HPIS sub-slots in the time slots. Please cite this article as: K. R. et al., Pre-channel scheduling and Priority-based reservation in medium access control for industrial wireless sensor network applications, Computers and Electrical Engineering (2017), http://dx.doi.org/10.1016/j.compeleceng.2017.05.010

JID: CAEE

ARTICLE IN PRESS K. R. et al. / Computers and Electrical Engineering 000 (2017) 1–15

…..

[m3Gsc;May 13, 2017;16:37] 5

…….

(a) HPIS

HPIS

…

…..

(b) Fig. 2. (a) general time slots and (b) HPIS-based time slots.

Fig. 3. Time slice boundaries.

The HPIS contains two sub-slots inserted in between the end of one slot to the start of the next slot. The deceleration of HPIS differentiates the high priority and low priority data to avoid the collision in traffic. The modified super frame structure with the HPIS is used to schedule the channel through the three processes as follows: core node selection, time scale boundary estimation and overlap mitigation. 3.1.1. Core node selection The intercommunication among the nodes using the multiple channels requires the temporal space in time slots of adjacent nodes. The separation of time slot into sub-slots (HPIS and data) is denoted as tHPIS and tdata . Based on the split-up of time slots, the identification of adjacent nodes and their scheduling is performed. The time slot assigned to the core node is reserved the space for priority estimation and normal periodic monitoring data. The selection of core node facilitates the communication between the adjacent nodes instead of sensor-to-sensor communication. This improves the data forwarding capability among the nodes and ensures the reliability of data transmission. Each node can have the different beacon interval (BI) and Superframe Duration (SD) in the communication system. In such situation, the configuration of core node is known as prior and their location is fixed. Once the core node is selected for transmission, the estimation of time slice boundary is the next stage of the scheduling process. 3.1.2. Time slice boundaries estimation Based on the cycle and the duration of superframe scheduling, the length of the time slice (TL ) is estimated. For each minor cycle (m), tHPIS starts at the time (mBImin ) and ends at the time (mBImin + Tm ). Then, tdata starts at (mBImin + Tm ) and ends at ((m + 1)BImin ) where the tHPIS starts. Fig. 3 illustrates the time slice boundaries for scheduling. 3.1.3. Overlap mitigation For each node, the channel scheduling algorithm proposed in this paper estimates the starting time and add this value to the initial minor cycle. Then, it periodically checks the minor cycles where the second time slice is larger than previous that fits the whole superframe duration for the selected node. The prior estimation of starting time and periodical evaluation of next minor cycle avoids the overlapping between the frames of different time slices. The Algorithm 1 listed following is used to perform the multi-channel scheduling: Initially, the HPIS is assigned to the time slots to provide the temporal separation between the current node and the core node. Then, the algorithm estimates the superframe duration and beacon interval for each node. Based on the BI level and the HPIS, the core node is selected. The initial scheduling comprises the checking whether the channel is free or not based on the slots. Then, the starting offset is assigned with the current instance of superframe and the corresponding beacon interval. The allocation of time intervals for tHPIS and tdata is performed with the ratio of beacon intervals (BImax , BImin ) and the scheduled beacon interval to the minor cycle. Please cite this article as: K. R. et al., Pre-channel scheduling and Priority-based reservation in medium access control for industrial wireless sensor network applications, Computers and Electrical Engineering (2017), http://dx.doi.org/10.1016/j.compeleceng.2017.05.010

ARTICLE IN PRESS

JID: CAEE 6

[m3Gsc;May 13, 2017;16:37]

K. R. et al. / Computers and Electrical Engineering 000 (2017) 1–15

Algorithm 1 1: Assign HPIS to the time slots 2: Estimate the superframe duration (SD), becon interval (BI) for each node 3: Select the core node (CN) //Scheduling in initial time 4: For each superframe Si 5: Select the free channel and add to the C 6: For each instance of superframe (j) 7: Assign starting offset: sij = jBIi 8: End For 9: End For //Update in scheduling 10: Compute the ratio of maximum beacon interval with the minimum beacon interval (BImax /BImin ) and assign to th 11: Form to (th − 1) 12: Set the minor cycle Tm = max(scheduled BImin to minor cycle) 13: tHPIS = mBImin to mBImin + Tm 14: tdata = mBImin + Tm to ((m + 1)BImin ) 15: End for

Fig. 4. Structure of data collected from the sensor node.

3.2. Multi-stage classification The creation of efficient data structures is the pre-requisite for classification of normal and emergency. The movement of complex operations into pre-processing stage acts as the base for efficient data structure creation. The equivalence table and the index table are the major parts of the data structure as shown in Fig. 4. The complexity lies in the traditional classification algorithms are a number of operations to find the intersection of bit vectors and the most searching operation for bit vector in the equivalent class table. The number of operations required to perform the comparison and intersection of N-bit vectors directly depends on the memory width (wp ) of the processor. Number of operation required to find the intersection = N/wp Number of operation required comparison = N/wp Besides, the length of equivalent table Tl also affects the searching of bit vectors in the table. Hence, the time complexity is defined by O(Tl N/wp ). The performance of comparison and searching operation requires further improvement based on the hash-based aggregation mechanism. The data collected by the core node from all the adjacent sensor nodes are converted into the bit vector form. The number of bit vectors is represented as B = b0 b1 b2 . . . bw p and the aggregated bit vectors of the input are defined byA = a0 a1 a2 . . . aw p . The ith bit of the array A is the aggregation of the chunk of input vectors only if the following bitwise operation is satisfied. w −1

p ai = ORi=1

(1)

Based on the Eq. (1), the aggregating process is repeated and finally all the bit vectors are aggregated into the single word with the tree structure as shown in Fig. 5. The mapping of N-bits in the packet header to the index with the certain classification rules is the major problem in classification. The algorithm comprises three phases such as phase 0, subsequent phase and the final phase. Each phase includes the memory lookup which is the reduction of bits requirement. Phase 0: The fields of packet header split-up into the multiple chunks, and they served as the index values to the multiple memories. Each lookup table provides the ID value calledEqID . Based on the ID value, we can identify the aggregated entries with 0 or 1. Subsequent Phase: The combination of results from the previous look-up constitutes the index value in the memory. Final Phase: The pre-computation of contents in the memory provide the value called Class_ID that specifies the status of data either normal or emergency. Storage optimization: Please cite this article as: K. R. et al., Pre-channel scheduling and Priority-based reservation in medium access control for industrial wireless sensor network applications, Computers and Electrical Engineering (2017), http://dx.doi.org/10.1016/j.compeleceng.2017.05.010

ARTICLE IN PRESS

JID: CAEE

[m3Gsc;May 13, 2017;16:37]

K. R. et al. / Computers and Electrical Engineering 000 (2017) 1–15

7

Fig. 5. Aggregated form.

Fig. 6. Reservation request frame.

The sequences in index table will increase the storage overhead if the node deployment is more. Once the simultaneous computation of sequences in the index table and the construction of data are over, the following processes are used to reduce the storage overhead. • • • •

Allocate the block of memory to data block and compute the EqID Store the unique ID values (α ) to the new table called jump tableJT. If α = 1, all the ID values of data are equal to the aggregated bit vectors If α < = ∝0 (∝0 = block size/4), the data set is compressed into two sets (equivalent and index) and the addresses are stored in the jump table • If α > ∝0 , no compressions are performed. The introduction of jump tables reduces the storage overhead considerably. 3.3. Reservation in slots The data classification through the recursive flow classification algorithm yields two types of data called normal and emergency. If the data is normal, then the data is forwarded to the fusion process and transmits the fused data to the memory directly. But, if the data are emergency means, the immediate response is required. Each node requires the additional sample data to compute the variations accurately. The amount of exceeding data called burst data streams. The arrival of burst streams at the same time due to the correlation between the nodes requires immediate priority to resolve it. The frame structure of reservation scheme is illustrated in Fig. 6. It includes three processes such as priority estimation, severity index and the volume [24]. 3.3.1. Priority estimation Based on the type of data measured by the sensor nodes in the industrial devices (voltage, current, power, temperature etc.,), the priority is determined. The node with the highest priority has the great privilege to access. 3.3.2. Severity index It is defined as the degree of deviation of data from the safety range and it is computed from following equation.

⎧  −V  ⎨min VVUL−V L SI = 0   ⎩ V −VU min

VU −VL

⎫

V ≤ VL ⎬ VL ≤ V ≤ VU ⎭ V ≥ VU

(2)

where, V – numerical value of collected data Please cite this article as: K. R. et al., Pre-channel scheduling and Priority-based reservation in medium access control for industrial wireless sensor network applications, Computers and Electrical Engineering (2017), http://dx.doi.org/10.1016/j.compeleceng.2017.05.010

ARTICLE IN PRESS

JID: CAEE 8

[m3Gsc;May 13, 2017;16:37]

K. R. et al. / Computers and Electrical Engineering 000 (2017) 1–15

VL – Lower Bound VU – Upper Bound The estimated SI from the Eq. (2) is within the range of 0 to 1. The higher values of SI denote the severe condition. 3.3.3. Volume prediction The additional slot allocation depends on the volume of the burst data in buffer. When the adjacent node receives the allocation request for frame, the node provides the reply ACK with the added time filed to inform the new frame structure illustrated in Fig. 6. The core node used in the network structure introduces the token to determine the order of nodes with the emergency data. The computation of token depends on the value of priority and severity index as follows:

token = ρ1 ×

P + ρ2 × SI 8

(3)

where, ρ 1 and ρ 2 – 0 to 1 with ρ 1 + ρ 2 = 1 The core node estimates the response from the adjacent nodes and if the token is high value, then the corresponding node has the privilege to transmit the data. Upon receiving the allocation request, core node validates whether each node assigns the additional time slots or not. If they assigned, then it forward the data to the sink to the adjacent nodes. Otherwise, it waits until the next superframe arrival. The nodes entered into the sleep interval if there are no burst data conserves the energy effectively. 3.4. Fusion The optimal solution for effective communication links is the basic requirement to validate the packet loss in the network. Due to the absence of prior notification of packet loss, the pre-processing stage by bounded message size is the difficult task. This problem initiates the prediction of bandwidth with the good estimation strategies. The operating phases involved in fusion processes are learning (based on measurement fusion) and expert phase (based on local filter estimates). 3.4.1. Learning The normal data received from the sensors are used to compute the Measurement Fusion (MF) are denoted as

xi = DNormali

Where, i = 1, 2, . . . .N

(4)

The core node (CN) estimates the best mean square estimator from the arrived data and they regarded as the fused data as follows:

xMF = E[DNormali ] i

(5)

The plants observed by the sensors are modeled as follows:

xt+1 = Axt + ut

(6)

yt = Ci xt + vti

(7)

where, ut and

vti

are the zero mean white Gaussian noises

A, C - coefficient matrices xt + 1 - Next state estimator yt – Column vectors The centralized matrix and the resultant column vector are defined by,

⎡

δ 1C1

⎤

⎡

δ 1 y1

⎤

C¯t = ⎣ ... ⎦; y¯t = ⎣ ... ⎦ δ NCN δ N yN

(8)

where, δ = binary random variable With the replacement of rows and columns by zero corresponding to the lost packets [25], the state estimate of MF is formulated as follows:





MF MF xˆt+1 = I − L¯ t C¯t Axˆt−1 + L¯ t y¯t





L¯ t = Pt−1C¯tT C¯t Pt−1C¯tT + R

(9) (10)

where, Pt − 1 – sending probability rate Please cite this article as: K. R. et al., Pre-channel scheduling and Priority-based reservation in medium access control for industrial wireless sensor network applications, Computers and Electrical Engineering (2017), http://dx.doi.org/10.1016/j.compeleceng.2017.05.010

ARTICLE IN PRESS

JID: CAEE

[m3Gsc;May 13, 2017;16:37]

K. R. et al. / Computers and Electrical Engineering 000 (2017) 1–15

9

Algorithm 2 1: 2: 2: 3: 4: 5: 6: 7: 8: 9: 10: 12: 13: 14: 15: 16: 17: 18:

Extract the classified output from the multi-stage classification (DNormal , DEmerg ) Collect the slots with tHPIS and tData If (D = DNormal ) // Fusion Compute the state estimate of measurement fusion by (9) Compute the probability estimate of MF by (12) If (information is loss) Compute the local filter estimates (LF) Compute the estimated fusion for each state variable by (14) Compute the probability of estimated fusion by (15) Estimate the optimal gain values of state estimates by (16) Endif Else // Reservation Assign the priority to the recent data from sensor nodes Calculate severity index and volume Compute the token to initiate the reservation request Reserve the slots for with priority space in tHPIS Reserve the slots for normal data tData Endif

The next state of probability with the MF measure is expressed as follows:

Pt+1 = APtMF AT + Q

(11)



−1 Where, PtMF = Pt−1 + C¯tT RC¯t



(12)

The utilization of MF in the learning phases brings two observations in data transmission as follows: • Recursive estimation of state estimates by using the inverse matrix form defined in (10), (11) and (12) by using the Moore-Penrose pseudo inverse with an assumption of R > 0 efficiently implements the Kalman filter. • If the packets received from sensors are the loss, then the information conveyed by the packet also loss during the learning phase. Hence, the expert phase is initiated to overcome such issue. 3.4.2. Expert On the basis of fusion of local filtered measurements, a new estimate is used to recover the information called Estimate Fusion (EF). The formulation of transmission of data from the ith node is expressed as

LFti = Oti LFt−1 + St i yti

(13)

where, LFlocal filtered version The core node utilizes the following fusion rule to compute the EF of normal data.

xˆtEF = E [xt ]LFt−1 =

N 

ϕi LFt−τi

(14)

i=1

where, LFt−τi – local recent filter estimate The corresponding probability estimation of the packet transmission rate is redefined as follows:



PtEF = var (xt ) − ϕt E LFt, τ LFt, τ T



ϕt T

(15)

This probabilistic value is used to evaluate the different choices for preprocessing stages and monitor the performance of on-line estimator effectively. The CN further estimates the optimal gain to reduce the computational complexity in the data transmission. The Open-loop Partial Estimate Fusion (OPEF) on the basis of latest received packet is expressed as

xt OPEF =

N 

ϕti,OPEF LFt−τi

(16)

i=1

With the measurements defined in (16), the computational complexity is reduced and the optimal performance is achieved with less noise measures. The Algorithm 2 is used to perform the reservation and fusion is listed as follows: The combination of reservation and data fusion listed in the above algorithm initially receives the time slots for transmission from the multi-channel scheduling. The line (2)–(9) shows the processes involved in fusion such as state and probability estimation for MF and EF. Similarly, line (12)–(18) shows the reservation process in time slots with three measures, namely, priority, severity level and the volume. The prior classification and the reservation in time slots for higher priority data increases the performance of network parameters such as throughput and packet delivery rate with less energy consumption and delay. Please cite this article as: K. R. et al., Pre-channel scheduling and Priority-based reservation in medium access control for industrial wireless sensor network applications, Computers and Electrical Engineering (2017), http://dx.doi.org/10.1016/j.compeleceng.2017.05.010

JID: CAEE 10

ARTICLE IN PRESS

[m3Gsc;May 13, 2017;16:37]

K. R. et al. / Computers and Electrical Engineering 000 (2017) 1–15 Table 2 Simulation parameters. Parameters

Values

Physical layer Network size CN location Number of nodes Data packet size Simulation period Listen interval Buffer size Initial energy Sample rate Bandwidth Data rate Data link layer Network layer Energy model

IEEE 802.15.4 50 × 50 m 25 m 100 4800 bits 150 s 10 ms 32 1J 10 samples / s 20 MHz 250 kbps Variable slot scheme Selective-repeat Residual energy

4. Performance analysis This section illustrates the validation of proposed MSC-RMAC algorithm against the existing algorithms of priority-based energy efficient MAC protocol with interval time (PRIN), Scheduled MAC and Timeout-MAC [9] real-time big data gathering (RTBDG) [15], and the energy-efficient connected target algorithms (AR-SC) [20] regarding the parameters of Packet Delivery Ratio, throughput, network lifetime and consumed energy. PRIN: The static priority assigning from the source node to the hop nodes in this scheme to achieve the QoS effectively. But, the enhancements in QoS requirements were required to fulfill the WSN capacity. The mathematical modeling of priority queues for prioritization assured the rapid delivery of sensitive and semi-sensitive packets. The two bits were added to the reduction of priority queues classified the priority levels effectively. SMAC: The major problems such as idle listening which caused the excessive energy consumption. Sensor-MAC (SMAC) periodically assigned the nodes in either sleep or active state to reduce the energy consumption. But, there is an energy consumption was observed during the listening period without any data transmission. TMAC: To reduce the excessive energy consumption, the Time-out MAC (TMAC) was proposed in which the nodes were periodically switched between active and idle state during the communication and listening period. The communication of a particular node with the neighboring nodes through the signals such as Request To Send (RTS), Clear To Send (CTS) and data Acknowledgement (ACK) provided the reliable communication. The hibernation of node in TMAC completely terminated the idle listening time per frame. RTBDG: The risk analysis in IWSN was investigated on the basis of the RTBDG algorithm with the data screening and clustering. The clustering was performed on the basis of the residual energy and Received Signal Strength Indicator (RSSI). The energy consumption was balanced in such approach for the specific operations. The proposed system was unsuitable if the application size was large. AR-SC: The adjustments in the sensing radius of each node on different power levels required the energy conservation. The low-power working environment allowed the sensor to operate for long-time period. Besides, the application of the ARSC method to adjust the sensing radius turned the network as stable. The node wish to join or leave has more power with the sensing radius adjustments. The simulation parameters to validate the proposed MAC algorithms in NS 2 simulator are listed in Table 2. 4.1. Packet delivery ratio The measure of the sum of packets received by the destination to the sum of packets generated is referred as packet delivery ratio. This section investigates the effect of the proposed work on PDR values with respect to minimum and maximum simulation intervals as shown in Table 3. The variations of PDR values against the simulation interval time (s) for different methods are shown in Fig. 7. In existing methods, the PRIN S provides high PDR values (82 and 95) for minimum and maximum simulation interval time due to the lack of pre-scheduling to forwarding. But, the prior to priority assignment, the MSC-based classification in proposed work increases the PDR to 86 and 96% of minimum (25) and maximum (150) time period respectively. The comparison between the proposed MSC-RMAC with the PRIN S shows that the proposed MSC-RMAC offers 4.65 and 1.04% improvement compared to PRIN S respectively. 4.2. Network lifetime The time horizon until the first node dies refers network lifetime. Network lifetime has the greatest impact on energy efficiency. Longer lifetime networks have the better performance in energy consumption. The variations of network lifetime Please cite this article as: K. R. et al., Pre-channel scheduling and Priority-based reservation in medium access control for industrial wireless sensor network applications, Computers and Electrical Engineering (2017), http://dx.doi.org/10.1016/j.compeleceng.2017.05.010

ARTICLE IN PRESS

JID: CAEE

[m3Gsc;May 13, 2017;16:37]

K. R. et al. / Computers and Electrical Engineering 000 (2017) 1–15

11

Table 3 Packet delivery ratio analysis. Simulation interval time (s)

Packet delivery ratio (%) SMAC

TMAC

PRIN S

AR-SC

RTBDG

MSC-RMAC

25 50 75 100 125 150

58 61 64 67 71 78

56 63 70 77 80 84

82 84 85 87 88 90

83 86 87 89 91 93

82 85 87 91 93 95

86 89 91 94 95 96

Fig. 7. Packet delivery ratio analysis.

Table 4 Network lifetime analysis. Number of rounds

Network lifetime (104 s) LEACH

HEED

WAIER

AR-SC

RTBDG

MSC-RMAC

1 2 3 4 5 6 7

2.014 2.142 2.018 1.89 2.14 1.698 1.784

2.50 2.24 2.63 2.31 2.14 2.30 2.41

2.65 2.13 2.24 2.04 2.43 2.25 2.35

2.30 2.05 2.37 2.36 2.42 1.86 2.03

2.126 2.214 2.34 2.104 1.985 2.014 2.0241

2.56 2.68 2.71 2.75 2.79 2.94 3.04

Fig. 8. Network lifetime analysis.

values with the simulation rounds (as shown in Table 4) depict the effectiveness of proposed MAC-RMAC over the existing methods. The variation of end-to-end delay values with respect to the number of rounds is shown in Fig. 8. For the higher simulation rounds, the lifetime for the network is high 2.41 × 104 s with HEED protocol. The preservation in bandwidth and the recovery of the packet loss via fusion technique reduces the energy consumption that increases the lifetime of the network in proposed MSC-RMAC. The lifetime of the network with the proposed MSC-RMAC is high 3.04 × 104 s which is 20.72% higher than the HEED protocol for higher rounds (7). Please cite this article as: K. R. et al., Pre-channel scheduling and Priority-based reservation in medium access control for industrial wireless sensor network applications, Computers and Electrical Engineering (2017), http://dx.doi.org/10.1016/j.compeleceng.2017.05.010

ARTICLE IN PRESS

JID: CAEE 12

[m3Gsc;May 13, 2017;16:37]

K. R. et al. / Computers and Electrical Engineering 000 (2017) 1–15 Table 5 Consumed Energy Analysis. Sample inter arrival rate

Consumed energy (mJ) SMAC

TMAC

PRIN S

AR-SC

RTBDG

MSC-RMAC

25 50 75 100 125 150 175 200 225

4.50 3.81 3.08 3.84 3.51 3.67 3.54 3.19 3.55

2.30 2.21 2.28 3.15 3.11 3.54 3.68 3.78 3.23

1.45 1.38 1.96 1.27 1.69 1.42 1.68 1.26 1.70

1.39 1.32 1.12 1.69 1.17 1.91 1.19 1.51 1.73

1.43 1.38 1.28 1.06 1.49 1.91 1.88 1.32 1.54

1.14 1.02 0.71 0.91 0.51 1.14 0.86 0.26 0.44

Fig. 9. Consumed energy analysis.

4.3. Consumed energy The total energy consumed by the proposed work during the transmission and reception depends on the power required for those processes as follows:

ET X = (PowerT X ∗ Packet size)/2(μJ )

(17)

ERX = (PowerRX ∗ Packet size)/2(μJ )

(18)

The sample inter arrival rate is linearly increased and the corresponding consumed energy in the network is noted in Table 5 and plotted as in Fig. 9. Due to the loss of packets, the data transmission in existing methods consumes more energy. Among them, the RTBDG algorithm offers less energy consumption (1.54 mJ) compared to others. The HPIS based time slot scheduling and the reservation of MAC time slots with the severity level and the volume in proposed MSC-RMAC reduces the energy consumption to 0.44 mJ. The comparison between the MSR-RMAC and RTBDG shows that the MSC-RMAC reduces the consumed energy by 71.43% compared to RTBDG for higher sample interval rate. 4.4. Throughput The number of data packets sent over the total simulation period refers throughput. The mathematical formulation for throughput is expressed as

T hroughput =

Number o f data packets sent (bits ) T ime period (secs )

(19)

The variation in simulation period and the corresponding throughput values are noted down in the Table 6 and plotted in Fig. 10. For the small simulation period, the throughput values of RTBDG are maximum compared to other existing methods of priority-based protocols. The effective data gathering mechanism in RTBDG provides high throughput values 100,994 and 94,757 bytes / s for small and higher simulation period. But, the reservation-based data transmission by a core node in proposed MSC-RMAC increases the throughput values of 3.54 and 15.21% compared to RTBDG respectively. Please cite this article as: K. R. et al., Pre-channel scheduling and Priority-based reservation in medium access control for industrial wireless sensor network applications, Computers and Electrical Engineering (2017), http://dx.doi.org/10.1016/j.compeleceng.2017.05.010

ARTICLE IN PRESS

JID: CAEE

[m3Gsc;May 13, 2017;16:37]

K. R. et al. / Computers and Electrical Engineering 000 (2017) 1–15

13

Table 6 Throughput Analysis. Simulation interval time (s)

Throughput (bytes/s) SMAC

TMAC

PRIN S

AR-SC

RTBDG

MSC-RMAC

25 50 75 100 125 150

89,803 85,572 89,203 89,733 89,809 87,969

84,911 86,988 85,553 86,903 86,431 85,403

84,046 84,410 86,252 83,778 86,402 86,972

90,074 90,710 89,953 90,835 89,872 90,522

100,994 93,745 93,392 103,892 104,318 94,757

104,704 106,288 105,449 108,373 103,600 111,756

Fig. 10. Throughput analysis. Table 7 End-to-end delay Analysis. Node density

30 40 50 60 70 80 90 100

End-to-end delay(ms) SMAC

TMAC

PRIN S

AR-SC

RTBDG

MSC-RMAC

6.32 8.16 9.68 11.41 13.09 14.77 16.45 20.36

4.24 5.98 8.24 10.15 12.15 14.15 16.15 18.15

2.36 3.42 4.16 5.11 6.01 6.91 7.81 8.71

2.13 3.02 3.56 4.33 5.05 5.76 6.48 7.19

2.11 2.02 3.55 3.77 4.41 5.74 6.15 6.24

1.98 1.38 3.05 3.53 4.32 4.89 5.23 5.39

Fig. 11. End-to-end delay analysis.

4.5. End-End delay The overall time duration for the data packets from the source node to destination node through the number of intermediate nodes refers end-to-end delay. The effectiveness of the proposed algorithm is determined by the minimum values of end-to-end delay. Table 7 shows the variations of end-to-end delay values with respect to the node density variations. From the Fig. 11, it is observed that the end-to-end delay values for the existing and proposed algorithms undergone the linear variations. For the small densities, the delay values of RTBDG are minimum compared to other existing methods SMAC Please cite this article as: K. R. et al., Pre-channel scheduling and Priority-based reservation in medium access control for industrial wireless sensor network applications, Computers and Electrical Engineering (2017), http://dx.doi.org/10.1016/j.compeleceng.2017.05.010

JID: CAEE 14

ARTICLE IN PRESS

[m3Gsc;May 13, 2017;16:37]

K. R. et al. / Computers and Electrical Engineering 000 (2017) 1–15

and TMAC. The effective data gathering mechanism in RTBDG provides minimum delay values 2.11 and 6.24 ms for small and higher node densities. But, the reservation-based data transmission by a core node in proposed MSC-RMAC reduces the delay values to 1.98 and 5.39 which is 6.16 and 13.6% reduction compared to RTBDG respectively. 5. Conclusion This paper discussed the issues in the satisfying of QoS requirements, scalability and security constraints. This paper proposed the reliable and energy efficient structure for reliable data transmission based on the fusion techniques. By performing the deep analysis of the characteristics of the data (normal and emergency), the request for resources is generated. The achievement of automatic adjustment of the percentage of sent messages by using the proposed MSC-RMAC is considered as an alternative way to increase the throughput and decrease the energy consumption. The selection of the core node from the group of sensor nodes, construction of HPIS on the basis of time slices and the boundary values decided the core node in the network. Based on the severity level, the data classification and the reservation of slots are performed to transfer the data to the sink node. The sequential processes in the MSC such as pre-processing, equivalent class set creation, index estimation and the storage optimization yields the normal and emergency data with minimum energy consumption. The major advantage of the proposed work is the provision of a trade-off between the network metrics, PDR, throughput and energy economy. The prior learning mechanism for the priority assignment provided the optimization in communication efficiency and the adjustment of sending rate under dynamic topologies. The experimental test bed is constructed to validate the performance of proposed MSC-RMAC against the existing protocols. The proposed MSC-RMAC provided the 1.04% improvement in PDR compared to existing PRIN S for higher simulation interval periods. Besides, the network lifetime is 20.72% higher than the HEED protocol. The throughput for MSC-RMAC also 15.21% higher by 71.43% less energy consumption compared to RTBDG method. The comparative analysis of the proposed algorithms with the existing protocols assured the significance of proposed energy efficient architecture in the industrial monitoring applications. References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25]

Akan OB, Karli OB, Ergul O. Cognitive radio sensor networks. IEEE Netw 2009:23. Akyildiz IF, Lee W-Y, Chowdhury KR. CRAHNs: cognitive radio ad hoc networks. AD Hoc Netw 2009;7:810–36. Chen R, Park J-M, Hou YT, Reed JH. Toward secure distributed spectrum sensing in cognitive radio networks. IEEE Commun Mag 2008:46. Bukhari SHR, Rehmani MH, Siraj S. A survey of channel bonding for wireless networks and guidelines of channel bonding for futuristic cognitive radio sensor networks. IEEE Commun Surv Tutorials 2016;18:924–48. Bukhari SHR, Siraj S, Rehmani MH. PRACB: a novel channel bonding algorithm for cognitive radio sensor networks. IEEE Access. 2016;4:6950–63. Zhang X, Luo Q, Cheng L, Wan Y, Song H, Yang Y. CRTRA: coloring route-tree based resource allocation algorithm for industrial wireless sensor networks. In: 2012 IEEE wireless communications and networking conference (WCNC). IEEE; 2012. p. 1870–5. Heinzelman WB, Chandrakasan AP, Balakrishnan H. An application-specific protocol architecture for wireless microsensor networks. IEEE Trans Wireless Commun 2002;1:660–70. Younis O, Fahmy S. HEED: a hybrid, energy-efficient, distributed clustering approach for ad Hoc sensor networks. IEEE Trans Mobile Comput 2004;3:366–79. Subramanian AK, Paramasivam I. PRIN: a priority-based energy efficient mac protocol for wireless sensor networks varying the sample inter-arrival time. Wireless Personal Commun 2017;92(3):863–81. Pantazis NA, Nikolidakis SA, Vergados DD. Energy-efficient routing protocols in wireless sensor networks: a survey. IEEE Commun Surv Tutorials 2013;15:551–91. Akhtar F, Rehmani MH. Energy replenishment using renewable and traditional energy resources for sustainable wireless sensor networks: a review. Renewable Sustainable Energy Rev 2015;45:769–84. Rashid B, Rehmani MH. Applications of wireless sensor networks for urban areas: a survey. J Netw Comput Appl 2016;60:192–219. Amjad M, Sharif M, Afzal MK, Kim SW. TinyOS-new trends, comparative views, and supported sensing applications: a review. IEEE Sens J 2016;16:2865–89. Hou L, Bergmann NW. Novel industrial wireless sensor networks for machine condition monitoring and fault diagnosis. IEEE Trans Instrum Meas 2012;61:2787–98. Ding X, Tian Y, Yu Y. A real-time big data gathering algorithm based on indoor wireless sensor networks for risk analysis of industrial operations. IEEE Trans Ind Inf 2016;12:1232–42. ˚ Zheng T, Gidlund M, Akerberg J. WirArb: a new mac protocol for time critical industrial wireless sensor network applications. IEEE Sens J 2016;16:2127–39. Suto K, Nishiyama H, Kato N, Huang C-W. An energy-efficient and delay-aware wireless computing system for industrial wireless sensor networks. IEEE Access 2015;3:1026–35. Anjum I, Alam N, Razzaque MA, Mehedi Hassan M, Alamri A. Traffic priority and load adaptive MAC protocol for QoS provisioning in body sensor networks. Int J Distrib Sens Netw 2013 2013. Yang D, Xu Y, Wang H, Zheng T, Zhang H, Zhang H, et al. Assignment of segmented slots enabling reliable real-time transmission in industrial wireless sensor networks. IEEE Trans Ind Electron 2015;62:3966–77. Han G, Liu L, Jiang J, Shu L, Hancke G. Analysis of energy-efficient connected target coverage algorithms for industrial wireless sensor networks. IEEE Trans Ind Inf 2015;13:135–43. Lu C, Saifullah A, Li B, Sha M, Gonzalez H, Gunatilaka D, et al. Real-time wireless sensor-actuator networks for industrial cyber-physical systems. Proc IEEE 2016;104:1013–24. Kreibich O, Neuzil J, Smid R. Quality-based multiple-sensor fusion in an industrial wireless sensor network for MCM. IEEE Trans Ind Electron 2014;61:4903–11. Pinto AR, Montez C, Araújo G, Vasques F, Portugal P. An approach to implement data fusion techniques in wireless sensor networks using genetic machine learning algorithms. Inf Fusion 2014;15:90–101. Yang L, Li C, Song Y, Yuan X. An energy-efficient 2R MAC based on IEEE 802.15. 6 for health monitoring. In: 2015 IEEE Globecom workshops (GC Wkshps). IEEE; 2015. p. 1–6. Schenato L. Optimal sensor fusion for distributed sensors subject to random delay and packet loss. In: 46th IEEE conference on decision and control, 2007. IEEE; 2007. p. 1547–52.

Please cite this article as: K. R. et al., Pre-channel scheduling and Priority-based reservation in medium access control for industrial wireless sensor network applications, Computers and Electrical Engineering (2017), http://dx.doi.org/10.1016/j.compeleceng.2017.05.010

JID: CAEE

ARTICLE IN PRESS K. R. et al. / Computers and Electrical Engineering 000 (2017) 1–15

[m3Gsc;May 13, 2017;16:37] 15

R. Kannan received B.E. degree in Electrical and Electronics Engineering from University of Madras, Chennai, India, in 1999; and M.Tech. degree in Power Electronics from VIT University, Vellore, India, in 2003. He is currently a Ph.D. Scholar in Faculty of Electrical Engineering, Anna University, Chennai, India. His research interests include Wireless Sensor Networks, Motor Fault Diagnosis, Power Electronics and Embedded Systems. S. Solai Manohar received his M.E degree in Applied Electronics from University of Madras, Chennai, India and Ph.D. degree in Electrical Engineering from Anna University, Chennai, India. Currently he is Professor in Department of Electrical and Electronics Engineering at KCG College of Technology, Chennai, India. His areas of interest are Embedded Networked Systems & Wireless Sensor Networks. M. Senthil Kumaran received M.E. degree in Applied Electronics from University of Madras, Chennai, India and Ph.D. degree in Electrical Engineering from Anna University, Chennai, India. Currently he is Associate Professor in Department of Electrical and Electronics Engineering at SSN College of Engineering, Chennai, India. His research interests include Power Electronics, Electrical Machines & Drives, Applied Electronics and Embedded Systems.

Please cite this article as: K. R. et al., Pre-channel scheduling and Priority-based reservation in medium access control for industrial wireless sensor network applications, Computers and Electrical Engineering (2017), http://dx.doi.org/10.1016/j.compeleceng.2017.05.010