Energy-efficient routing protocol for reliable low-latency Internet of Things in oil and gas pipeline monitoring

Sensing nodes and master nodes characteristics

SNs are responsible for sensing data and transmitting it to MN. However, MN can do both sensing and gathering of data from its nodes and then transmit it to the upper layer. To make a low-cost and energy-efficient network, simple SNs are enabled with short-range communication whereas, the range of MNs is made adaptive to communicate with UAVs. Additionally, MNs are deployed at a hop distance of ten nodes from each other. Therefore, the number of SNs is ten times of MNs in the network (Mahbub & Shubair, 2023).

Functional architecture of IoT nodes

An IoT sensing node in comprised of the following units: sensing, location finding unit, processing unit, transceiver, battery, and memory, temperature sensor, pressure sensor, ultrasonic sensor for corrosion and erosion, and flow sensor: special piezo sensors for highly accurate flow measurements are embedded in the sensing unit of the IoT node.

Energy consumption model

IoT nodes use energy while transmitting and receiving messages. To analyze the different data delivery models, we must define the energy needed for transmitting and receiving messages in the IoTs network. For this purpose, we use the same radio-energy dissipation model as the one used by Gupta & Shekokar (2016) and Anand et al. (2016). Energy and distance between the source and destination node are directly proportional to each other. The smaller the distance, the lesser the energy consumed, and vice versa. However, there is a defined threshold distance between transmitting and receiving nodes given by Eq. (1): (1)${\displaystyle d_o=\sqrt{E_{fs}/E_{mp}}}$

where E_fs is the free space energy and E_mp is the multipath energy. If node A wants to forward a message to node B and their distance is more than this threshold distance d_o, then the energy required to send or receive a data packet is given by Eq. (2). (2)${\displaystyle E_{con}=\left(PL\times E_{tx}\right)+\left(E_{mp}\times PL\times {\left(d_{AB}\right)}^4\right)}$

If the distance between nodes is less than the specified threshold, the communication is happening in the near-field and the energy consumed can be calculated by Eq. (3). (3)${\displaystyle E_{con}=\left(PL\times E_{tx}\right)+\left(E_{mp}\times PL\times {\left(d_{AB}\right)}^2\right)}$

where E_con is the energy consumed, PL is the Packet Length, E_tx is the transmission energy, d_AB defines the distance between nodes A and B.

IoTs system working model

The detailed functioning model of IoTs is presented in this section. Based on the importance of the message, this layer has several objectives to achieve.

• The primary focus revolves around optimizing energy efficiency while transmitting regular OGI parameter values to ensure the network’s long-term sustainability.

• The paramount goal shifts to achieving low latency for critical data values.

• Fault tolerance and reliability with minimal data loss is a significant concern in both aforementioned scenarios.

To achieve all these objectives, we first design a basic mechanism for network deployment and route discovery, complemented by modules dedicated to enhancing reliability and fault tolerance. In conclusion, we present a suite of corresponding algorithms that encapsulate and facilitate these diverse endeavors.

Important message-based routing protocol (IMRP) for (Ω = 0)

Let us consider a network with N sensing nodes, where N ∈ {1, 2, 3, …, N}, and M master nodes, where M ∈ {1, 2, 3, …, M}. It’s important to note that the number of master nodes (M) is always less than the total number of sensing nodes (N), specifically M < N. When Ω = 0, data transmission occurs at uniform time intervals. The entire time duration is segmented into discrete time slots denoted by t, where t ∈ T and T ∈ {0, 5, 10, 15, 20, …}. In instances where a sensing node (SN) is unable to complete transmission within the allotted time, data loss might occur due to buffer overflow, as sensing nodes are constrained by limited buffer capacity.

Step 1: After network deployment and link establishment phase, individual nodes acquire their respective locations and dual routes. These routes encompass a path to master node 1 and another to master node 2, each accompanied by a dedicated route table containing details of h_c and nh.

Step 2: Let’s consider the scenario where the ith sensing node aims to transmit m_i bits of sensory data with Ω = 0 to the Master Node within a time frame of T seconds. This endeavor entails the following scenarios:

Scenario 1: both neighbors operational: In this scenario, both routes of Node i possess sufficient energy reserves for data transmission. In such instances, Node i evaluates both routes and opts for the one with a lower h_c value leading to MN. To enhance network reliability and energy efficiency, a mechanism is in place to detect nodes approaching energy depletion, denoted by energy levels falling below a predefined threshold γ. In such cases, the node enters a sleep mode, briefly reactivating solely for transmission within a limited timeframe. Additionally, it communicates with its neighboring nodes, indicating its availability for a duration of t_x, while simultaneously notifying the master node of its energy status for necessary interventions. The selection of the next hop for Scenario #1 is visually depicted in Fig. 1.

Scenario 2: single available route: In this scenario, if the preferred route is inaccessible due to energy depletion of the next-hop node (nh), Node i opts for route 2 for data transmission. To optimize energy consumption during transmission, Node i disseminates information to all nodes along route 2, notifying them of route 1’s unavailability. Subsequently, any node wishing to transmit data will directly opt for route 2, and vice versa. The operational mechanism is visually depicted in Fig. 2.

 
____________________________ 
Algorithm 1 Important Message-based Routing Protocol (IMRP)_________________________________________ 
 1:  t ← 0 
 2:  while t ⁄= 0 do 
 3:       while alive_nodes ⁄= 0 do 
 4:           Find alive_nodes 
 5:           packets_transmitted ← packets_transmitted + 1 
 6:           if En ≤ 0.05 then 
 7:                Add node to dead node 
 8:           end if 
 9:           DATA(a,b,c) Determine datatype 
10:           if Ω = 0 then 
11:                if mod(t,10)==0 then 
12:                     if route1 is available then 
          ROUTE1(nh) 
           Add (pl ∗ (Etx)) to (Emp ∗ pl ∗ nnd(SN,NN)2) 
           Subtract it from En 
13:                     else if route2 is available then 
          ROUTE2(nh) 
           Add (pl ∗ (Etx)) to (Emp ∗ pl ∗ nnd(SN,NN)2) 
           Subtract it from En 
14:                     else 
          Buffer data 
           Wait for UAV and transmit 
           Add (pl ∗ (Etx)) to (Emp ∗ pl ∗ nnd(SN,NN)4) 
           Subtract it from En 
15:                     end if 
16:                end if 
17:           else 
18:                if route1 is available then 
          ROUTE1(nh) 
           Add (pl ∗ (Etx)) to (Emp ∗ pl ∗ nnd(SN,NN)2) 
           Subtract it from En 
19:                else 
          Increase Rc 
20:                     if nodeb/wSN&MNisalive then 
          Transmit data to that node 
           Add (pl ∗ (Etx)) to (Emp ∗ pl ∗ nnd(SN,NN)4) 
           Subtract it from En 
21:                     else 
          Transmit data to MN 
           Add (pl ∗ (Etx)) to (Emp ∗ pl ∗ nnd(SN,NN)4) 
           Subtract it from En 
22:                     end if  
23:                end if 
24:           end if 
25:       end while 
26:       t ← t + 5 
27:  end while________________________________________________________________________________________________________    

Scenario 3: both routes unavailable: When both routes are unavailable for data transmission, Node i enters a listening mode and stores data in its buffer for a span of n readings. Subsequently, it awaits the arrival of upper-layer UAVs within range, extending its communication range as necessary. This facilitates data exchange and the dissemination of information regarding the inoperable neighboring nodes for necessary actions as shown in Fig. 3.

Important message-based routing protocol (IMRP) for (Ω = 1)

In situations where the sensed packet type is deemed critical, meaning it falls outside the predefined normal temperature and pressure ranges, low latency data transmission becomes imperative. This urgency is aimed at mitigating potential risks to both lives and assets. To address such exigencies, we outline a series of scenarios employing distinct mechanisms:

Scenario #1: Both neighbors operational: It is the same as in the previous routing mechanism described for Ω = 0. If both routes are alive, the route with less hop count is considered for critical packet transmission to reach the master node as soon as possible. If the node has the remaining energy equivalent to threshold energy γ, it continues its transmission without delaying toward the nearest master node. Moreover, it notifies the master node about the energy status of the sending node.

Scenario #2: Single available route: if the route with less hop count is not available i.e., the route to the nearest master node is unavailable. In that case, the node will increase its transmission range and send data to the next-to-next-hop neighbor on the preferred route. In case no node between the nearest master node and the sending node is available, the node will send that critical data directly to the master node. Next hop selection mechanism for scenario 2, Ω = 1 is illustrated in Fig. 4.

Scenario #3: Both routes unavailable: For critical messages, the prime objective is to send data to the master node in minimum time to reduce loss. If both the neighbors are dead for transmission, the node will increase its range and transmit the data to an alive node between the master node and the sending node. In case all nodes in between are dead, sending node will directly transmit data to the master node or if a UAV from the upper layer is in range, it will transmit a message through the UAV as shown in Fig. 5. The Algorithm 1 describes the functionality of IMRP

 
_________________________________________________________________________________________________________________________ 
Algorithm 2 Packet Optimization for Important Message based Routing Protocol (PO-IMRP)___ 
 1:  t ← 0 
 2:  while t ⁄= 0 do 
 3:       while alive_nodes ⁄= 0 do 
 4:           Find alive_nodes 
 5:           packets_transmitted ← packets_transmitted + 1 
 6:           if En ≤ 0.05 then 
 7:                Add node to dead node 
 8:           end if 
 9:           DATA(a,b,c) Determine datatype 
10:           if Ω = 0 then 
11:                if mod(t,10)==0 then 
12:                     if route1 is available then 
          ROUTE1(nh) 
           Add (pl ∗ (Etx)) to (Emp ∗ pl ∗ nnd(SN,NN)2) 
           Subtract it from En 
           Attach data part to packet and forward 
13:                     else if route2 is available then 
          ROUTE2(nh) 
           Add (pl ∗ (Etx)) to (Emp ∗ pl ∗ nnd(SN,NN)2) 
           Subtract it from En 
           Attach data part to packet and forward 
14:                     else 
          Buffer data 
           Wait for UAV and transmit 
           Add (pl ∗ (Etx)) to (Emp ∗ pl ∗ nnd(SN,NN)4) 
           Subtract it from En 
15:                     end if 
16:                end if 
17:           else 
18:                if route1 is available then 
          ROUTE1(nh) 
           Add (pl ∗ (Etx)) to (Emp ∗ pl ∗ nnd(SN,NN)2) 
           Subtract it from En 
           Attach data part to packet and forward 
19:                else 
          Increase Rc 
20:                     if nodeb/wSN&MNisalive then 
          Transmit data to that node 
           Add (pl ∗ (Etx)) to (Emp ∗ pl ∗ nnd(SN,NN)4) 
           Subtract it from En 
           Attach data part to packet and forward 
21:                     else  
          Transmit data to MN 
           Add (pl ∗ (Etx)) to (Emp ∗ pl ∗ nnd(SN,NN)4) 
           Subtract it from En 
22:                     end if 
23:                end if 
24:           end if 
25:       end while 
26:       t ← t + 5 
27:  end while________________________________________________________________________________________________________    

Packet optimization for important message-based routing protocol (PO-IMRP)

Data forwarding in industrial IoTs whose nodes are powered by batteries is significantly hampered by energy consumption. These nodes expend a lot of energy forwarding, receiving, and temporarily storing messages. In OGI, the critical messages must reach master nodes for quick actions. We have proposed Packet Optimization for an Important Message Routing Protocol (PO-IMRP) which significantly reduces energy consumption and increases the lifetime of the network by optimizing data packets.

Sensor nodes deployed on pipelines are constantly sensing and reporting readings to master nodes at equal time intervals. In PO-IMRP, once a particular node has some parameter readings to transmit, it selects the route and collects the data from all the nodes residing on the route. IoT node packets have two parts: header and data part. The header contains the info of ID_n, MN, h_c, nh, and route status. The data part contains the values of temperature, pressure, and flow rate. However, all the sensed data is heading toward master nodes. Neighbor nodes can combine their data with the packets for further transmission.

Let node A intend to transmit packet p_id1 to MN, it selects the route with a lower h_c, represented by node B. Once the packet reaches node B, which holds its own data packet p_id2 for transmission, an optimization is employed. Rather than transmitting both packets individually, each with distinct headers, node B consolidates the data and prepares a new packet featuring a unified header. Given the shared destination for both packets, this approach significantly conserves energy that would have been expended in transmitting two separate packets with distinct headers. The scenario where nodes A and B possess distinct packets for transmission is depicted in Fig. 6. An illustrative example showcasing the merging of a data packet with an identical header at the subsequent hop is presented in Fig. 7. The operational procedure is outlined in Algorithm 2 .

PO-IMRP packet format

The optimized packet format and its size are illustrated in Fig. 8. The whole packet has two parts: Header and Data. The header part is comprised of five fields i.e., node ID, master node ID, hop-count, next-hop, and route status. Each field contains 1 byte and a total of 5 bytes. The Data part combines the data with one separator field. The first three fields are for the first data packet comprised of 6 bytes then a separator field with 1 byte and then again data fields (6 bytes). The packet size increases on each hop.

Simulations setup

The setup for simulations consists of simple sensing nodes (SN) and master nodes (MN). Simple sensing nodes are the units that contain temperature, pressure, and flow rate sensors along with the transceivers. These units have some predefined sensing and communication ranges respectively. However, master nodes have the same characteristics as simple sensing nodes but with an adaptive communication range. In our architectural framework, SN transmits data to MN, which serve as sinks. Typically, the distance from sensor nodes to the sink should fall within the range of 0 to 200 m. Consequently, we position MNs at a distance of 100 m. The simulation field we use to represent the OGI pipeline is 100 × 5 meters with MNs at both ends. The sensing range R_s of IoT nodes is set to 4.5 m and the communication range R_c to 10 m. Given our previous discussion that SNs are deployed at a distance of $\sqrt{3}\ast R_s$, each SN is positioned 8 m apart. SN sense data at uniform time intervals i.e., t is set to 5s. while the total time is segmented into discrete time slots i.e., T is set to 10s. The initial energy E_init for each IoT module is set to be 0.5 J. Data transmission and reception energies E_t and E_r is 10⁻⁶ J. Free-space Energy E_fs is set to 10⁻¹¹ J, multipath Energy E_mp as 1.3∗10⁻¹⁵ J, and data aggregation Energy E_aggr as 5∗10⁻⁹ J. The parameters used in simulations are outlined in Table 3. Node deployment is visually illustrated in Fig. 9.

In our study, the sensing nodes SNs acquire parameter values at uniform time intervals, denoted as t seconds. When these sensed values fall within the predefined ranges for temperature, pressure, and flow rate, data transmission adheres to the discrete-time slot T. In cases where the sensed values deviate from the specified ranges, triggering the identification of critical data, immediate transmission ensues, bypassing the designated time slots. Consequently, within the context of monitoring OGI pipelines, two distinct data categories emerge: normal and critical. To cater to the importance of these data types, we have introduced two routing mechanisms, namely IMRP and PO-IMRP. Within this section, we carry out an energy comparison of these proposed techniques.

Energy consumption vs packets transmitted

In this section, we will perform a comparative analysis of the energy consumption associated with both proposed routing protocols, IMRP and PO-IMRP. E_t signifies the energy utilized for packet transmission, while E_r represents the energy used during packet reception.

For PO-IMRP

In the PO-IMRP protocol, the packet size is adaptive and not fixed. When a node initiates transmission, it gathers and appends the data portions of all the other nodes through which it hops toward the master node (MN). Initially, the packet length is 11 bytes, consisting of 5 bytes for the header and 6 bytes for the data segment. However, with each hop from one node to another, the packet length grows by 7 bytes, encompassing 6 data bytes from the current node’s data and an additional 1-byte separator. Therefore, initially,

PL = 11 bytes

With each successive hop, the packet length (PL) undergoes an update of 7 + PL. (9)${\displaystyle PL=7+PL.}$

Let’s calculate the total energy consumption for a single round for PO-IMRP.

First half energy consumption PO-IMRP: The energy consumption rule for ith node will be:

$E_i=\left[i{E}_t+i{E}_r\right]\cdot \left(7j+PL\right)$

where, i ∈ [1, 6] and j ∈ [0, 5]

For,

Node 1: i = 1j =0 $⟹E_1=\left[1E_t+1E_r\right]\cdot \left(7\left(0\right)+PL\right)$

Hence, (10)${\displaystyle E_1=\left(E_t+E_r\right)\cdot PL}$ Put E_t = E_r = E and PL = pl₁ in Eq. (10) (11)${\displaystyle E_1=2E\cdot {pl}_1}$ Similarly, (12)${\displaystyle E_2=\left(2E_t+2E_r\right)\cdot \left(7+PL\right)}$ Put E_t = E_r = E and (7 + PL) = pl₂ in Eq. (12) (13)${\displaystyle E_2=4E\cdot {pl}_2}$ Hence, ${\displaystyle \text{Node 3:}{E}_3=\left(3E_t+3E_r\right)\cdot \left(14+PL\right)⟹E_3=6E\cdot {pl}_3\dots \dots }$ ${\displaystyle \text{Node 6:}{E}_6=\left(6E_t+6E_r\right)\cdot \left(35+PL\right)⟹E_6=12E\cdot {pl}_6}$

Therefore, the total energy consumption for the first half is: ${\displaystyle E_{\text{first-half}}=\sum _{i=1}^6{E}_i{E}_{\text{first-half}}=E_1+E_2+E_3+\dots +E_6}$

E_first−half =  [2E⋅pl₁] + [4E⋅pl₂] + [6E⋅pl₃] + [8E⋅pl₄] + [10E⋅pl₅] + [12E⋅pl₆] (14)${\displaystyle E_{first-half}=E\left(2{pl}_1+4{pl}_2+6{pl}_3+8{pl}_4+10{pl}_5+12{pl}_6\right)}$

Second half energy consumption PO-IMRP: Similar to the first half, the second half of IoT nodes’ energy consumption will be the same, i.e., (15)${\displaystyle E_{second-half}=E\left(2{pl}_1+4{pl}_2+6{pl}_3+8{pl}_4+10{pl}_5+12{pl}_6\right)}$

Total energy consumption PO-IMRP The total Energy consumption can be calculated as:

$E_{total}={\sum }_{i=1}^2{E}_i$

Adding Eqs. (14) and (15)

E_total = [E(2pl₁ + 4pl₂ + 6pl₃ + 8pl₄ + 10pl₅ + 12pl₆)] + [E(2pl₁ + 4pl₂ + 6pl₃ + 8pl₄ + 10pl₅ + 12pl₆)] (16)${\displaystyle E_{total}=E\left({pl}_1+2{pl}_2+3{pl}_3+4{pl}_2+5{pl}_5+6{pl}_6\right)}$

Packets transmitted-PO-IMRP: In the PO-IMRP protocol, the packet length is not a fixed value; instead, it adapts to optimize packet transmission. When an IoT node initiates data transmission, it appends the data segment of each subsequent hopped IoT node, leading to an increased packet length at each hop until it reaches the destination. Consequently, the number of packets transmitted in a single round is higher compared to IMRP. To determine the total number of packets transmitted in a single round, consider the following: Node 1 initiates the process by transmitting a single packet with a length of 11 bytes to MN. As Node 2 transitions to the transmission phase, it dispatches an 11-byte packet to Node 1. Node 1 then appends additional sensed values to the packet before forwarding the extended packet to MN. Consequently, this process enables the transmission of two sets of data without causing network congestion i.e., an 11-byte packet by Node 2 and an 18-byte packet by Node 1 is transmitted to MN in a single transmission. This pattern repeats for the remaining nodes. A summarized representation of packets and transmissions for the entire round can be found in Table 5.

Packets congestion control management

Our proposed IoTs routing mechanism, PO-IMRP offers packet control management by optimizing packets. Table 6 presents the total number of transmissions, rounds, and packets delivered by IMRP and PO-IMRP. IMRP accomplished 942 rounds and facilitated the transmission of 11,194 packets throughout the network’s lifetime. In contrast, PO-IMRP generated 426 packets while efficiently transmitting 17,784 data values through optimized packets. Through a comprehensive analysis of these statistics, it can be concluded that PO-IMRP performs well by delivering more data in fewer rounds and generating fewer packets in the network.

Effect on transmissions per round (IMRP vs PO-IMRP)

Figure 10 shows the relationship between transmissions and rounds. The peak values of both show that PO-IMRP performs more transmissions in fewer rounds than IMRP.

Effect on energy consumption

Figure 11 describes the pattern between net energy of the network per round throughout the network lifetime. Although the energy consumption of PO-IMRP is more than IMRP the transmission in the network increases significantly. The net energies of IMRP and PO-IMRP after performing transmissions of round 1 are as shown in the Table 7. Hence, the Energy gap is calculated as: (18)${\displaystyle E_{gap}=5.9962-5.9932=0.003\mathrm{J}}$

Performance analysis of PO-IMRP vs LEACH

We have performed a performance analysis of our proposed protocol PO-IMRP with LEACH. Low Energy Adaptive Clustering Hierarchy (LEACH) is one of the popular WSN routing protocols. It is proposed by Heintzelman. The idea behind choosing LEACH for comparison is that it is a renowned WSN routing protocol, capable of enhancing the lifetime of the network, and provides fault tolerance Mohapatra & Rath (2019). Our analysis involved implementing LEACH within the context of our linear WSN setup and associated parameters, enabling a comprehensive comparison of outcomes.

Operational nodes per transmission (PO-IMRP vs LEACH)

The relationship between operating nodes per round is shown in Fig. 12. In PO-IMRP, the first node dies after 392 rounds and 16,464 packets have been transmitted. However, in LEACH, the first node dies between 600–700 rounds and less than 1,000 packets have been transmitted. This indicates that PO-IMRP shows more transmissions than LEACH.

Table 7:

Remaining energies of nodes for a single round.

Node	Energy-IMRP (J)	Energy PO-IMRP (J)
1	0.4995	0.4988
2	0.4996	0.4991
3	0.4996	0.4994
4	0.4997	0.4996
5	0.4998	0.4998
6	0.4999	0.4999
7	0.4999	0.4999
8	0.4998	0.4998
9	0.4997	0.4996
10	0.4996	0.4994
11	0.4996	0.4991
12	0.4995	0.4988
Total energy	5.9962	5.9932

DOI: 10.7717/peerjcs.1908/table-7

Energy consumption per transmission:

Fig. 13 depicts the correlation between Energy and transmissions of the IMRP, PO-IMRP, and LEACH protocol. LEACH shows energy consumption between 0.01 to 0.22 J. However, PO-IMRP shows an average energy consumption of 7.066 × 10⁻³ J. This analysis demonstrates that PO-IMRP is more energy efficient than LEACH.