Sage Journals: Discover world-class research

Abstract

With the diversification of industrial Internet of Things applications, there is a growing demand for mobility support in industrial wireless networking environments. However, the routing protocol for low-power and lossy networks is designed based on a static environment and is vulnerable in a mobility environment. Routing protocol for low-power and lossy networks is an Internet engineering task force standard in the low-power and lossy network environments used mainly in industrial environments. In addition, although routing protocol for low-power and lossy networks is based on collection tree protocol and is suitable for data collection and upward traffic transmission, it struggles with downward traffic transmission in terms of control, actuation, and end-to-end transmission. In this article, the problems caused by mobile nodes in routing protocol for low-power and lossy networks are discussed, and a retransmission scheme named IM-RPL is proposed. This retransmission scheme can improve the performance of downward traffic for the mobile nodes by retransmitting the packets to the neighbor nodes, the mobile node’s new parent sets, and relaying them to the mobile node. Its performance is evaluated through an experiment. The results demonstrate that using OpenMote in OpenWSN’s time slotted channel hopping induces a packet reception ratio improvement and a lower transmission delay as compared to standard routing protocol for low power and lossy.

Keywords

Routing protocol for low-power and lossy networks mobility Internet of Things wireless sensor networks downward traffic 6tisch

Introduction

Starting with Industry 4.0, many technologies are gaining attention for improving manufacturing plants, such as industrial IoT, Cyber Physical Systems (CPS), and Digital Twin. These technologies aim to improve productivity and competitiveness through the optimization of processes, reduction of defect rate, and monitoring of a manufacturing environment, by combining information and communication (ICT) technology with manufacturing industry–related fields.¹

These techniques are commonly based on data collected from a large number of sensors installed in industrial sites. Industrial wireless communications technology is a critical technology used to connect a large number of sensors in a new plant paradigm (e.g. with small batch production, mobile sensing, and smart devices for operators).²

To satisfy the quality of service (QoS) requirement according to the diversification of industrial IoT applications, there is a constant demand for connectivity and mobility among devices in industrial environments such as smart grids and automation systems.³ A standard operation procedure support system is provided via radio equipment for field worker patrols and maintenance. A reliable braking system for an automated guided vehicle (AGV), a maintenance period determination system operating by analyzing device behavior patterns, a system for wireless connection of a hard-wired environment, and applications for factory automation all require connectivity and mobility support in various industrial environments. Thus, wireless communication technology is required.

Efforts to apply the wireless communication technology represented by a wireless sensor network (WSN) in the industry have been made for approximately 10 years. Consequently, standard technologies such as WirelessHART and ISA100.11a have been established;⁴ IEEE 802.15.4e⁵ time slotted channel hopping (TSCH) MAC technology has been used to secure network reliability and communicate with industrial devices. However, with the recent emergence of industrial IoT, the need for IP network connection has emerged,⁶ and various working groups (WG) of IETF are actively working toward this purpose.

Routing over low-power and lossy (ROLL) networks WG defines the network environment in which low-power devices operate as a low-power and lossy network (LLN). LLN refers to a network in which embedded devices with limited power, memory, and processing resources are connected via links such as IEEE 802.15.4, Bluetooth, low power, and Wi-Fi. As LLN’s communication equipment is connected by a series of multi-hop radio links, ad hoc on-demand distance vector (AODV) routing and dynamic source routing (DSR), which are standardized by the IETF’s mobile ad hoc network (MANET) WG, can be used as routing protocols for LLN. However, the abilities of these routing protocols in power, memory, and process are limited. Wireless links also have lower power and higher loss characteristics than those considered by MANET.⁷ Therefore, it is difficult to apply MANET routing protocols to LLN.

To overcome this problem, ROLL proposed RPL, a routing protocol that can operate flexibly in network environments with unreliable connections. RPL is a de facto standard for LLN environments that provides multi-hop connectivity for many battery-powered, embedded, wireless devices for data transmission and communication.⁸

However, as RPL is designed based on a static environment, mobility is a challenge it needs to address. In addition, RPL is suitable for collecting data based on Collection Tree Protocol (CTP).⁹ In the case of upward traffic for data collection, the root node can be reached with the best route by transmitting the packet to the upper node. The downward traffic is less reliable than the upward traffic because of the slow updation of the route.¹⁰ Furthermore, in an environment containing mobility, the reliability of downward traffic is significantly lowered. However, because low latency performance and determinism must be guaranteed in an industrial environment,¹¹ the part of the system that improves the reliability of downward traffic such as P2P communication, advanced metering infrastructure (AMI) setting, and activation is also an important issue.¹²

In this article, the problems caused by mobile nodes (MNs) in an RPL environment are discussed, and a retransmission scheme named IM-RPL is proposed. This scheme comprises a technique to increase the reliability of the downward traffic in the presence of mobility and a quick parent-change method to support a dynamic environment of frequent changes. Experiments were conducted in OpenWSN’s TSCH-based environment using OpenMote. The proposed scheme exhibits a higher reliability and lower transmission delay than RPL.

The composition of this article is as follows. In section “Background,” the background of RPL is introduced. In section “Related work,” recent researches on upward/downward traffic and the mobility of RPL are introduced. In section “Downward Traffic Problems and Mobility Support Technique,” the problem of mobile and downward traffic in the standard RPL is discussed, and mobility support and a quick parent-change scheme is proposed. In section “Experiment,” the results of real experiments and simulations in the OpenWSN environment with an OpenMote board based on CC2538 are discussed. Section “Conclusion and future work” describes the conclusion and future work.

Background

This section includes the study of TSCH and RPL.

802.15.4e: TSCH

TSCH is a MAC protocol that focuses on providing reliable wireless communications in industrial environments, such as in process control and monitoring automation. IEEE 802.15.4e is an extension concept in 802.15.4 that minimizes transmission delay with a non-contention-based MAC protocol, and it provides high reliability in a low-power environment. TSCH supports multi-hop and multi-channel communication over application domains such as industrial automation and process control through time-division multiple access (TDMA). TSCH is a combination of time slot access, multiple channels, and channel hopping. Time slot access prevents collisions between competing nodes, increases packet throughput, and provides deterministic latency to applications. Multiple channels allow different nodes to communicate in the same slot frame by using different channel offsets. This increases network throughput. Channel hopping also improves communication reliability by mitigating the effects of interference and multipath fading. Thus, TSCH maintains a very low duty cycle through time slot access; increases energy efficiency, network capacity, and reliability; and provides predictable latency. TSCH is used in topologies such as star, tree, and mesh. It enables the efficient use of resources through frequency hopping and is suitable for multi-hop networks.¹³

RPL

RPL is designed for low-power and high-noise network environments such as IEEE 802.15.4 and powerline communication. It supports various routing metrics to accommodate the diverse requirements of various applications. For this purpose, the functions corresponding to routing metrics and path optimization, to achieve the requirements of each application, are separated by an objective function (OF). In general, the OF may include the expected transmission count (ETX), power consumption, and received signal strength indication.

Constructing DODAG for routing

RPL aims to construct a direction-oriented directed acyclic graph (DODAG) topology, and routing is based on the topology. The DAG root denotes the root of a non-circulating graph with directionality, and DODAG is a DAG root with one destination. DODAG is built on the DODAG information object (DIO), destination advertisement object (DAO), and DODAG information solicitation (DIS) control messages. Although the upward traffic routing paths such as multi-to-point are built using the DIO, routing is required for the downward traffic such as control, extraction, end-to-end, and messages from external LLN networks. Information for the downward traffic is communicated via DAO messages.

The DODAG configuration process is as follows. The root node periodically transmits broadcast information in a DIO message required for building a DODAG topology. A DIO message is used to establish an upward traffic path, set up a parent, and send traffic to the root to the parent. This allows neighboring nodes to participate in the network and updates the topology. The child node selects the best parents through the DIO message and sends the path information in the DAO messages to the root to construct the downward traffic path. The root node contains routing information for the entire node and establishes a downward path. DIS messages are control messages sent to receive the DIO messages, and they are used when a node first joins the network or needs to receive frequent DIO messages, such as in the case of MNs.

Storing/non-storing mode

RPL has two modes: storing and non-storing. In non-storing mode, the intermediate node does not store the routing table, and the root node manages the path for the entire downward traffic. If the packet transmission to the neighboring node is not one-hop distance, it is transmitted to the root and reaches the destination via the source routing. In storing mode, the intermediate node stores the routing table and contains the routing information of the lower nodes of the node. The node sends the DAO to the parent, and each parent node possesses its own routing table. The parent node receives the DAO and sets the path to the routing table. Subsequently, the DAO is sent again to the parent to modify the parent table. Each node has information regarding its subordinate nodes. Both modes have their advantages and disadvantages. If many nodes are participating in the network, there may be a shortage of memory, as significant resources are required for the routing table in storing mode. However, bottlenecks for nodes near the root can be reduced as compared to the non-storing mode.

Related work

This section includes recent research on upward/downward traffic and mobility issues of TSCH and RPL.

Upward/downward traffic

Upward/downward traffic issues are one of the most actively researched areas in RPL. In an industrial sensor network environment where data are mainly collected from the sensor nodes through wired/wireless communication, RPL based on CTP is the most suitable routing protocol. Recently, there have been steady efforts to increase the reliability of sensor data collection. Experiments on sensor information monitoring, upward traffic environment construction, and performance evaluation are underway in the RPL.^14–16 However, so far, research has been focused on objective selection and functions for constructing DODAG.^17–24 A study of improving connectivity between the parent node and the child node is much required.

Although RPL has strengths in upward traffic, it is unreliable in downward traffic such as in control and P2P communication. Research is needed to improve the reliability of downward traffic. Studying upward/downward traffic issues can improve the performance of bidirectional data, and such studies have been conducted.^25,26

Five-Nines²⁵ introduced a gradient metric to meet the reliability requirements (error recovery time, failure rate) required in various industrial environments and improve accurate link measurement and downward traffic in real time.

DT-RPL²⁶ identifies the problems in RPL path configuration and suggests a method to update the link quality between the parent node and the child node through upward/downward traffic. The proposed scheme improves the downward traffic rate, as well as the upward traffic. Despite any improvements in the reliability of upward/downward traffic, the packet transmission rate may drop sharply if there is mobility in the RPL.

Mobility

The RPL standard is designed to be static, so it is challenging to support mobility. When the routing node is moved, there are significant overheads in addition to low node fault, global repair, and overall network performance degradation. In recent years, however, there has been a growing demand for connectivity and mobility between operators and devices.²⁷ As applications diversify, research is needed to support a dynamic environment in which both static and mobile environments exist to ensure QoS for users.

ME-RPL²⁸ presents a mechanism for supporting mobility in the static environment of the standard RPL. The degree of mobility is determined according to the number of parent changes and the DIS message period is adjusted accordingly. In addition, if mobile and static nodes are candidate nodes for the preferred parent, a static node is selected as a parent to prevent frequent topology changes.

MoMoRo²⁹ supports mobility in the LLN network by introducing MoMoRo, a mobile support layer that can be easily applied to existing data collection protocols. MoMoRo periodically collects the information of adjacent nodes and evaluates link quality by using a fuzzy estimator. This estimator continuously updates the threshold for determining the fuzzy set, so it can easily adapt to the dynamic channel environment.

Corona RPL³⁰ uses the Corona mechanism to improve RPL. It creates and manages all neighbor nodes with IDs that have a minimum hop distance from the LoWPAN Border Router (LBR). The main objective is to ensure QoS while providing better mobility support to RPL. However, excessive use of control traffic in the Corona RPL can reduce the overall network performance, and it can vary greatly depending on the network size.

MRPL³¹ continuously checks the connectivity from the received signal strength indicator (RSSI) and improves the connectivity with the MN by transmitting the DIS message when the quality of the link is poor. In addition, the adoption of the smart-hop algorithm and the implementation of additional timers reduce the handoff delay of mobility.

RL-Probe³² monitors link quality by using reinforcement-learning techniques based on the multi-armed bandit model. RL-Probe combines synchronous and asynchronous monitoring mechanisms to maintain up-to-date information on link quality. Through this, a node can detect the movement of the MN and respond quickly.

In Sanshi and Jaidhar,³³ the authors have proposed a technique for quickly updating preferred parents by applying multiple routing metrics to the OF and maintaining additional parent tables. The multiple routing metrics include the distance between nodes based on the RSSI.

GTM-RPL³⁴ creates a cooperative game where nodes compete for network resources. This finds the best solution to maximize gain and minimize cost of the entire network. A node monitors the mobility of the node based on the RSSI value received and considers the priority and current noise level. The mobility metric and density metric are used to derive the mobility cost function and the energy cost function.

EMA-RPL³⁵ is a technique that delegates mobility processing processes to an associated node (AN) which is a preferred parent of a MN, reducing energy consumption of MNs and providing a seamless connection. The previous AN of the MN predicts the new AN. The connection with the previous one is disconnected only after the predicted AN and the MN is connected

Although much research has been conducted on the importance of mobility in RPL, there is insufficient research to improve the reliability of downward traffic in an RPL environment with mobility. Most techniques focus on how to select a preferred parent, but there is a lack of research on downward packet loss when a MN is moving.

Downward traffic problems and mobility support technique

This section discusses the upward/downward traffic problems and mobility problems in an RPL environment, and it proposes a suitable method to address them.

Upward/downward problem

RPL is a three-layer protocol based on CTP designed to be suitable for upward traffic, such as in sensing data collection. For upward path configuration, each node continuously measures the quality of the links to its neighboring nodes and reconstructs the path by selecting the best parent. It is also highly reliable, because each node can forward traffic to the root node simply by sending packets to their parent node. However, if the destination of downward traffic is not a neighbor node, the packet is directed to the root node, which manages the entire network routing table, and the downward traffic is passed to the destination through the source routing. Unstable nodes that can induce invalid routing information or topology changes such as the selection of a new parent create packet loss. This persists until a DAO message that updates the downward routing information arrives at the root node.

In addition, unlike upward traffic mechanisms, for downward traffic from parents to child nodes, unstable links or sustained packet loss do not affect parent changes. In other words, link measurements and parent changes through downward traffic are not made, even in situations where the quality of the link between nodes is unstable and the traffic is often lost. In that regard, Figure 1(a) shows the PRR for up/down traffic in the RPL, and it presents the average ratio of reception of end-to-end up/down traffic measured for each of the 6 h using a 10 OpenMote board. The ratio was measured at 91% for downward traffic and 96% for upward traffic.

Figure 1.

(a) Upward/downward traffic PRR in static network and (b) upward/downward traffic PRR of mobile node in dynamic network.

Mobility problem

As RPL is designed on a static basis and does not account for mobility, a technique is required to handle mobility. Figure 1(b) shows the upward/downward traffic PRR of a MN in RPL. The 10 OpenMote board was used to measure the reception rate of packets destined for the MN in the presence of mobility. The PRRs of the upward/downward traffic in the mobile environment were 55% and 45%, respectively. These are significantly lower than those in the static environment. As a node moves away from the previous parent’s transmission range, packet loss increases before it joins the new parent and network.

In RPL, an MN cannot act as a routing node and acts only as a terminal. It also does not mention the periodicity of the DIS message, which is essential in the presence of a MN in a standard document. RPL also uses a trickle timer³⁶ for DIO message transmission, which makes it suitable for static environments. If the node confirms that there is no change in topology, the period of the DIO message increases rapidly. Conversely, the period is initialized when the topology change is confirmed. However, the use of the trickle timer in a dynamic environment that requires fast participation of MNs increases the average transmission period of the DIO message and slows down the overall topology updation.

In RPL, a node undergoes the process of selecting a new parent after receiving a DIO message. If the rank of the neighboring parent list is smaller than the current parent rank and the difference between the two values is larger than the designated threshold value, the parent is reselected or changed. If the rank difference is smaller than the threshold value, the parent is not changed, because there is a high possibility that the overhead that occurs during the change is greater than the benefit of the parent change. However, it is difficult to organically support an environment in which fast parent changes are required.

Figure 2 shows the DODAG configuration and the routing path when there is downward traffic in a node with mobility in a non-storing mode. When node C sends a packet to mobile node D, it forwards the packet to the root node for routing. The root node checks the destination information of the packet in its routing table and performs source routing to set the complete path. The packet of node C arrives at node D via the A-B nodes. In a static environment, there is not much loss for downward packets, but problems arise when moving with mobility characteristics, such as for mobile node D.

Figure 2.

Example of downward traffic routing in non-storing mode.

Figure 3 depicts the process of downward traffic loss and path recovery after moving the MN. As the mobile node D moves, it cannot receive the packet of node B because node D is outside the transmission range of node B, which is the previous parent. The routing information recorded for the mobile node D in the network becomes invalid, creating packet loss. This occurs until the MN connects with a new parent through a control message and before the DODAG configuration and network participation.

Figure 3.

Downward traffic loss and path recovery process after mobile node moves.

In this proposed scheme, additional services to improve the downward traffic reliability to the MN are provided.

The contributions of the proposed scheme, IM-RPL, are as follows:

It increases the reliability of downward traffic and reduces transmission delay through a retransmission scheme for a MN in a mobility environment.

It supports rapid change of the parent and path recovery of the MN by sensing mobility and transmitting additional information to the MN.

The role of the MN in IM-RPL is defined as a terminal node as in the standard RPL, and it cannot serve as a routing node.

Downward traffic support scheme under mobility

The MN sends a DAO message to the root for the downward traffic route after selecting the new parent. Figure 4 displays the timing diagram of the DODAG participation of the MN. Mobile node D periodically sends a DIS to the nodes to select a new parent. The MN receives the DIO from neighboring nodes in response to the DIS, selects a new preferred parent, and updates the route by transmitting a DAO message to the root node. At this point, the mobile node D adds a mobility field in the DAO message, sends it to the root after selecting the parent node, and informs the parent node that it has mobility.

Figure 4.

Timing diagram of when mobile node participates in DODAG.

Figure 5 shows the DAO control messages and the DAO message options in RFC 6550. DAO messages can pass valid options. As shown in Figure 5, Mobility_flag is added to the DAO option field to inform the parent node that a node has mobility. The parent node checks the mobility of the child node from the DAO message and stores the mobility information in the mobility table.

Figure 5.

DAO option field and Mobility_flag.

Algorithm 1 shows that a parent node receives a DAO message from its child and then stores the mobility information in a mobility table through the mobility flag check. This can provide additional services for the downward traffic to the MN in the future.When a MN moves based on its locality, the new parent of the MN has a neighbor relationship with the previous parent.

Algorithm 1. Set mobility phase
1:	procedure DAO_MESSAGE handler
2:	if(DAO_MESSAGE→Mobility_flag = 1) then
3:	Set mobility table(Mobile_node_ip, set)
4:	end if
5:	end procedure

Figure 6(a) shows that the mobile node D receives downward traffic while it selects a new parent and updates its route. If node B, the previous parent of mobile node D, receives a downward packet destined for node D, it checks whether the child node is an MN. Because node D is mobile, node B broadcasts the downward packet to the neighbor nodes. Node E, the neighbor node receiving the broadcast message, can forward the packet to the MN because it knows the location information of the destination node D from the DIS message periodically transmitted by the MN.

Figure 6.

(a) Downward routing of the proposed scheme and (b) downward traffic path after path recovery.

Algorithm 2 is a downward traffic retransmission algorithm for the MN. When the downward traffic is directed to the MN, if the parent node does not receive the ACK for the packet destined for the MN, the set number of transmissions is decreased by 1. Failure to receive an ACK for a specified number of transmitted packets may lead to the inference that a child node with mobility has moved to another location. In other words, if ACK from a node that is registered in the mobility table is not received for the designated number of transmissions, it is determined that the mobile child node has moved. The downward packet’s next-hop MAC address is changed to a broadcast address, and the packet is retransmitted.

Algorithm 2. Retransmission phase
1:	procedure Downward_packet_sent_noack
2:	if(packet→NumOfTransmission = 0) then
3:	if(packet→next hop is in the mobility table) then
4:	packet→next_hop_mac_addr ← Broadcast
5:	Retransmission(packet)
6:	end if
7:	Else
8:	packet→Mobile_NumOfTransmission—
9:	Retransmission(packet)
10:	end if
11:	end procedure

It is important to note that the new parent of the MN is very likely to have a neighbor relationship with the parent of the previous MN. The broadcasted packet is transmitted to the nodes in the new parent list of the MN. The nodes receiving the DIS message from the MN know the location information of the MN. Among these, the nodes receiving the broadcasted packet forward the packet to the MN.

This technique reduces the packet loss by allowing the MN to receive the packet, even when the MN leaves the previous parent, selects a new parent, and restores the path.

Quick parent-change support scheme

The following scheme is proposed to solve the problem of loss of downward traffic until a MN selects a new parent and restores a path. In RPL, ETX is generally used as an OF. The ETX update to the neighboring node is mainly possible in the presence of the upward traffic to the neighboring node. The ETX calculation and RANK value are updated when sent to nodes, such as in control messages and generic packets. Based on the updated ETX and the RANK of the DIO message of the neighboring node, the optimal parent is selected and the DODAG topology is updated. RPL is based on CTP and is suitable for data collection. The reliability of upward traffic is higher than that of downward traffic. As the child node always selects the parent based on itself, the link update from the child node to the parent is frequent and accurate, whereas the link between the child nodes viewed from the parent node may be slow or inaccurate. When the link from the child node to the current parent node is identified as unstable, it is possible to quickly update the link and locate a new parent. However, even if there is a loss of downward packets from the parent node to the child node, it does not affect the change of the parent of the child. In addition, even if there is traffic to a neighboring node, a lack of persistence will make the change of the parent not work well.

When the difference between the RANK value of the current parent and the new best parent candidate does not exceed a certain threshold value, the parent node is not changed. Therefore, many downward packet losses are generated in a dynamic environment in which a fast parent change occurs.

To solve the above problems, a quick parent-change method is proposed. When the previous parent node of the MN uses the technique proposed in section “Downward traffic support scheme under mobility,” it sends 1-bit information in the packet to broadcast. Figure 7 shows the 802.15.4 format, including the flow control field of the MAC header. In the flow control field of the MAC header of the broadcasted packet, the information sent to the 1 bit can be used to immediately change the parent with the MN that receives the packet.

Figure 7.

Reserved field in 802.15.4 Mac frame format.

Figure 8 shows the control message processing and the parent selection process of the MN. The MN periodically broadcasts a DIS message to request the neighboring nodes for DIO messages and provides its own location information. However, the MN does not function as a routing node. If the MN acts as a routing node or a parent node, movement of the MN results in a network fault for the lower node. The MN is allowed to perform the role of the terminal node alone to avoid any confusion in the network. In the RPL, the MN does not function as a router, but only as the terminal node. Therefore, the DIO message is not generated and transmitted to the neighboring node.

Figure 8.

Process of control message and parent selection in mobile node.

The MN receiving the DIO message modifies the parent list based on the information and updates the rank through the OF. Subsequently, the MN undergoes a process to select a new parent. If the rank value of the best parent candidate is smaller than the rank value of the current parent, and if the difference between the two ranks is greater than a specified threshold value, the MN reselects and changes the parent. Specifying the threshold value in parent selection prevents frequent parent changes. The overhead incurred by changing the parent may be larger than when it is not changed.

However, the existing method is inappropriate for a MN that must quickly reselect a parent and update the route. The MN receiving the packet from the neighboring node checks a reserved_bit of the flow control field of the MAC header to proceed with the change of the parent. In this case, the threshold value is not considered, and the node whose rank value is best among the parent list is selected as the parent. This is because the MNs need to make parent changes faster to reduce invalid routing information while maintaining the link. When the parent changes, the DAO message is immediately sent to the root node. The root node that has a routing table for the entire topology receives the DAO message and modifies the route. This technique improves the reliability of downward traffic, reduces the delay time, and supports fast path recovery through rapid parent selection in a dynamic environment where mobility exists in RPL.

Analysis of packet overhead

Network packet overhead in a mobile environment varies greatly depending on the moving direction and speed of the MNs, making quantitative analysis difficult. Therefore, for packet overhead analysis, it is assumed that the MNs are disconnected at a constant rate from their preferred parent and receives downward packets at regular intervals.

When an MN receives $d$ downward packets for $t$ time, the MN receives one downward packet every $t / d$ time $(d \geq 1)$ . The MN disconnects from its preferred parent, causing $m$ downward packets to go the wrong previous path for $t$ hours. The total number of MNs in the network is $N_{MN}$ .

The wrong downward path to the MN is restored to the correct path by the MN sending a DAO message. In the $NETWOR K_{standard}$ which is a standard RPL network without IM-RPL implementation, it takes up to DAO cycles $(Perio d_{DAO})$ for the wrong downward path to be correctly recovered. If $t / d$ is less than $Perio d_{DAO}$ , then up to $PeriodDAO / (t / d)$ downward packets will be lost before path recovery. However, in IM-RPL implemented network $NETWOR K_{IM - RPL}$ , downward packets sent through the wrong path are also forwarded to the MN using the technique discussed in section “Downward traffic support scheme under mobility,” and the path is quickly restored using the technique detailed in section “Quick parent-change support scheme.” The MN will send a DAO message for path recovery within $t / d$ time even if the preferred parent of the MN is disconnected. Therefore, $m_{im - rpl}$ which is $m$ value in $NETWOR K_{IM - RPL}$ has a smaller value than $m_{standard}$ which is $m$ value in $NETWOR K_{standard}$

$m_{standard} > m_{im - rpl}$ (1)

The number of downward packets sent to an MN over time $t$ in $NETWOR K_{standard}$ is $d$ . Here, implementing IM-RPL and using the technique discussed in section “Downward traffic support scheme under mobility,” the number of downward packets to one MN in $t$ time

$d + (m_{im - rpl} \times (1 + N B_{prevP} \cap N B_{MN}))$ (2)

where $N B_{prevP}$ is a neighbor nodes set of the MN’s parent node and $N B_{MN}$ is a neighbor nodes set of the MN when the MN is disconnected from its parent. $N B_{prevP} \cap N B_{MN}$ is a set of nodes that are neighbors of the previous parent node of the MN and at the same time are neighbors of the MN. The nodes can deliver the downward packet transmitted to the wrong path to the MN. In equation (2), $1 + N B_{prevP} \cap N B_{MN}$ is the sum of one broadcasting packet transmitted by the previous parent and the number of packets $N B_{prevP} \cap N B_{MN}$ to be transmitted to the MN by the neighboring node of the previous parent.

In addition, the technique detailed in section “Quick parent-change support scheme” incurs the DAO packet overhead of the MN. Number of DAO packets that one MN makes over time $t$ in $NETWOR K_{IM - RPL}$

$NDA O_{standard} + m_{IM - RPL}$ (3)

where $NDA O_{standard}$ is the number of DAO packets made by an MN in $Networ k_{standard}$ .

Since the MN disconnected from its parent node is not neighbor with the parent node, $NBprevParent \cap NBmn$ is a small value, and $m_{im - rpl}$ is smaller than $m_{standard}$ , so it can show high reliability with little additional overhead.

Experiment

In this study, the TSCH MAC protocol-based 802.15.4e with high reliability in industrial low-power and congested environments is tested. In the case of CSMA/CA, which is one of the industrial IoT protocols, it is difficult to secure reliability in environments where industry-specific characteristics exist, such as increases in the number of nodes, interference, avoidance, multi-channel, multi-hop, and low power. TSCH avoids collision between competing nodes, increases packet throughput, and guarantees performance in multi-channel, multi-hop environments through channel hopping. In this study, experiments have been conducted using hardware devices and a simulator in the OpenWSN³⁷ environment.

As mentioned in section “Related work,” there are many techniques for dealing with metrics for parent changes, neighbor monitoring, and upward packet reliability in a mobile environment. However, there is a lack of research that focus on the loss of downward traffic in mobile environments. Therefore, in this experiment, which measures the performance related to downward packets, comparisons with other techniques that focus on upward traffic will not make much sense. As an alternative, we compare only two schemes, standard RPL and IM-RPL.

Experiment using hardware devices

The hardware device used is OpenMote,³⁸ which is showed in Figure 9, and it is based on CC2538.

Figure 9.

OpenMote used in the experiment.

The topology is configured as shown in Figure 10 and has distances of two hops, three hops, and four hops from the MN to the root node. The MN moves around the static node, which is the farthest from the root node. The topology includes one root node, one transmission node (packet generator), one MN (packet receiver), and five static nodes. Node A is the root node, nodes B to G are the static nodes, and node H is the MN.

Figure 10.

Topology_1, topology_2, and topology_3 in experiment using hardware devices.

The experiment environment parameters are shown in Table 1. In the verification of the experiment, the MN must deviate from the transmission range of the parent node. However, as the transmission range of CC2538-based OpenMote is 50–100 m, this is a significant challenge. For the mobility experiment, a 15 dBm attenuator is installed in each node to reduce the transmission range by 30 dBm attenuation. In the experimental scenario, Node A generates downward packets, transmits packets to the MN every 3 s, and does so for 5 min. Through experiments, the packet reception rate and transmission delay of the MN are confirmed.

Table 1.

Experiment environment parameters.

Parameter	Value
Experiment environment	OpenWSN
Mac protocol	TSCH
Mote type	OpenMote
Traffic periodicity	3 s
Mobile node speed	0.24 m/s
Attenuator	15 dBm
Experiment time	5 min

WSN: wireless sensor network; TSCH: time slotted channel hopping.

Figure 11 shows the receiving signal strength value according to OpenMote’s transmission distance and is derived from the Friis technique, a free-space radio loss model. The theoretical values were verified using a tool distributed by TEXAS INSTRUMENTS. It indicates the receiving signal strength indicator (RSSI) value for the distance used by the CC2538-250 kbps communication chip and is the theoretical value for the TX power of 0 and −30 dBm, respectively. The tool from TEXAS INSTRUMENTS is based on an environment that uses 2.4 GHz bandwidth and satisfies line-of-sight (LOS). Figure 12 shows the measurement of the receiving strength values according to the transmission distance of OpenMote after 0 and 15 dBm attenuation on each node. The measurements were taken by using two OpenMote units based on CC2538, and the RSSI values were measured based on the distance of the nodes. Figures 11(a) and 12(a) show that if the TX power is 0 dBm, the maximum limit distance for receiving packets is 80 m from the theoretical value and 50 m from the measured value. Figures 11(b) and 12(b) show that if the TX power is −30 dBm, the limit distance for receiving packets is 4 m from the theoretical value and 4 m from the measured value. If the distance exceeds the limit distance, each node cannot recognize the packet. MNs are assumed to move objects in the industrial environment, such as an AGV, a mobile workbench, and workers, at the speed of 3 m/s. The experiment environment was constructed with a speed of 0.24 m/s at a 4 m range, assuming the mobility of the vehicle of 3 m/s at a 50 m range.

Figure 11.

Theoretical value of RSSI according to transmission distance of OpenMote: (a) Tx power under 0 dBm attenuation and (b) Tx power under 30 dBm attenuation.

Figure 12.

Experimental RSSI value according to OpenMote transmission distance: (a) Tx power under 0 dBm attenuation and (b) Tx power under 30 dBm attenuation.

Resynchronization and routing update delay

TSCH requires synchronization of inter-node for communication, and it is established by a beacon message. The control messages such as DIS, DIO, and DAO that construct the DODAG managed by the network layer are based on this synchronization. There is no problem in selecting a new parent under synchronization according to the movement of the MN or in participating in the network. However, if it is in an asynchronous state, communication is not possible, and it is synchronized after receiving the beacon message.

In TSCH, the channel is continuously changed for each timeslot via channel hopping. In the asynchronous state, the channel for receiving the enhanced beacon message also changes continuously, resulting in a long time to synchronize. This interferes with establishing communication between nodes and increases latency. To reduce the resynchronization time, the time of synchronization is reduced by fixing the standby channel of the enhanced beacon in the asynchronous mode. In addition, in the OpenWSN environment, a DAO message is sent periodically through the timer. However, choosing a new parent and immediately sending a DAO message to the root node can lead to an update of the fast topology routing table. To do this, the DAO message is set to be transmitted immediately after the parent change.

Figure 13 is a graph of the average latency from the MN’s asynchronous state in the OpenWSN until it joins the network and changes the routing path of the root. The proposed method is applied to OpenWSN, and the time from desynchronization to synchronization is reduced from 26 to 15 s.

Figure 13.

Average latency from the asynchronous state to route update of mobile node.

Performance evaluation

The proposed method can perform differently depending on the deployment of the topology. Therefore, three different topologies were experimented with, as shown in Figure 10. Topology_1, topology_2, and topology_3 have two, three, and four hops, respectively, from the MN to the root. The movement path is as shown in Figure 10. Figure 14 shows the end-to-end packet reception ratio (PRR) of the MN according to the hop distance and the topology. IM-RPL (the proposed scheme) is more reliable than standard RPL because it can receive packets through neighboring nodes, even if the MN exits the transmission range of the previous parent. In the RPL, when the MN is outside the transmission range of the parent and cannot receive downward traffic, a new parent is selected through the control message, and packet loss occurs until the routing path is updated. As the number of hops increases, the distance from the MN to the route increases, and the packet reception rate is slightly reduced. The difference between the average PRR for the proposed method and RPL is approximately 20.

Figure 14.

End-to-end packet reception ratio of mobile node according to topology.

Figure 15 shows the end-to-end transmission latency of the MN after moving it. The transmission delay time is measured after the MN leaves the transmission range of the parent and until the next packet is received. The proposed scheme can receive the packet from the neighboring node through the downward traffic retransmission method even if the MN exits the transmission range of the parent node. Thus, it is confirmed that the transmission delay time of the proposed scheme is less than 4 s. For the standard RPL, if the MN does not receive the downward traffic outside the transmission range of the parent, the MN reassigns the parent through the control message, and the transmission delay occurs until the routing path is updated. Figure 15 shows that an average transmission delay of 16 s occurs.

Figure 15.

End-to-end transmission latency of the mobile node.

Figure 16 is a graph of the PRR according to the moving speed of the MN. In the proposed scheme, the moving speed of the MN is an important factor. To apply the proposed scheme, the MN must select a new parent and transfer the mobility information to the parent. The speed limit is the time that guarantees the entire update time of the routing path after exchanging control messages with the neighbor node, selecting a new parent, joining the network, and transmitting the DAO to the root node as the node moves. When the speed of the MN exceeds the speed limit, it is difficult to apply the proposed method, and the PRR drops sharply. In other words, a problem occurs because the node speed is higher than the time required for network participation. In this experiment, the transmission range of the node is 4 m, and the average time required to update the path by participating in the new network is 16 s. Therefore, a problem occurs when the MN moves by more than 4 m within 16 s. In the experiment, the limit speed is set to 0.25 m/s, which is 4 m/16 s, and the PRR sharply decreases when the speed exceeds the limit speed. In Figure 16, it can be seen that when the MN is slower than the limit speed, the reliability of IM-RPL is higher than that of standard RPL owing to the application of the proposed scheme. However, the PRR is drastically reduced when the MN is over limit speed. As the speed increases, the PRR of the proposed method and RPL converge to similar values, and it is observed that the proposed scheme does not affect the PRR of the MN. In addition, as the quick parent-change method is applied, the MN receives the packet from the neighboring node. It checks the MAC header control field of the packet and immediately changes the parent having the highest rank among the parent list for supporting fast path recovery. It is confirmed that when the quick parent-change method is added, the reliability of the PRR is higher than that of the IM-RPL.

Figure 16.

Packets reception ratio according to mobile node’s movement speed.

Overhead evaluation

Overhead comparison experiments were conducted to investigate the influence of control messages and retransmission messages on network performance. For broadcast messages for DIO, DIS, and retransmission, the TSCH transmits in a contention-based manner on the first slot (share) of the superframe. The shared slot is a slot for transmitting beacon messages, keep-alive messages, and control messages for network establishment. Therefore, the increase of the number of control messages and broadcast messages cause a collision between the nodes in the shared slot and a high probability of transmission failure, and it adversely affects information transmission and topology updation. In addition, although the DAO message is transmitted from a dedicated cell, the increase of the control messages may increase the total traffic in the network, thereby deteriorating the overall performance. In the experiment, the control message period is 60 s for the DAO message, 30 s for the DIO message, and 10 s for the DIS message of the MN. The experiment was based on topology_2 in Figure 10.

Figure 17 shows the number of control messages generated by the RPL. The node comprises eight nodes, from node A to node H, including a root node, a one-hop distance node from the root, a two-hop distance node from the root, and a MN. Node A periodically generates only DIO messages as a root node, and 10 DIO messages were transmitted to neighboring nodes for 5 min. Node A manages the entire routing path as the root node and does not generate the DAO message for updating the downward routing information. In the case of nodes B and C, which are located one-hop away from the root node, there is no change in the topology, and DAO messages are transmitted once every 60 s and five times in the experiment. Nodes D, E, F, and G located two hops away from the root node are affected by the DIS message of the MN and propagate the DIO message to the neighboring node more frequently than the existing period. According to the experimental path, nodes E and F, which are frequently exposed to the transmission range of the MN, receive more DIS messages than nodes such as D and G deployed at the end of the topology. On average, it can be seen that the nodes of 2-hop distance transmit the DIO messages 15 times more compared to the 10 DIO messages transmitted in case of the 1-hop distance. As nodes A through G are assumed as static nodes, they do not generate DIS messages. Mobile node H only acts as a terminal node; it does not generate a DIO message. The MN periodically sends a DIS message to the neighboring node to identify a new parent and update the path. In the experiment, one DIS message was generated every 10 s. During the experiment, 30 DIS messages were generated. As node H moves and continuously changes the new parent, it generates a DAO message and transmits it to the root node. The MN generates more DIS messages and DAO messages than other nodes.

Figure 17.

Number of control messages in RPL.

Figure 18 presents the number of control messages and retransmission messages in the IM-RPL. The number of control messages from the A to G nodes is similar to that of RPL. The broadcast message for retransmission of the proposed scheme occurs at nodes D, E, F, and G, which are the parent nodes of mobile node H. The average number of broadcast relay messages of 2-hop distance nodes is seven. In case of a DAO message, IM-RPL created eight more DAO messages with frequent parent changes than the standard RPL. This was done by using the fast parent-change support technique and addressing OpenWSN problems such as synchronization delay.

Figure 18.

Number of control messages and retransmission messages in IM-RPL.

Table 2 shows the number of control messages and the rate of change in the case of RPL and the proposed scheme. The message increase rate is −0.7%, and there is almost no change. Only the MN generates DIS messages, and the total number of generated messages was 30. The number of DAO messages in the proposed scheme increased by 18% more than RPL. The total number of control messages in the experiment was 203 in the RPL and 210 in the IM-RPL, and the total message increase ratio was 3%.

Table 2.

Number of control messages and increase rate.

	RPL	IM-RPL	Increase rate (%)
Number of DIO	130	129	−0.7
Number of DIS	30	30	0
Number of DAO	43	51	18
Total	203	210	3

RPL: routing protocol for low-power and lossy networks; DIO: DODAG information object; DIS: DODAG information solicitation; DAO: destination advertisement object.

Table 3 lists the number of retransmission messages for downward traffic in standard RPL and the proposed scheme. The retransmission message can be changed according to the number of retransmissions of Algorithm 2. The number of retransmissions is set to 1 in the experiment. In the proposed scheme, there are broadcast overheads for transmitting downward traffic to neighbor nodes, and 31 additional broadcast retransmissions occur as compared to RPL.

Table 3.

Number of retransmission messages.

	RPL	IM-RPL
Number of message	None	31

RPL: routing protocol for low-power and lossy networks.

According to the 802.15.4e standard, each slot in the TSCH is 10 ms. In this experiment, the slotframe consists of 9 slots and has a length of 90 ms per superframe. The superframe has one shared slot and eight dedicated slots, which can be adjusted as per user requirements. Broadcast messages such as Beacon, DIS, DIO, and retransmission messages are transmitted in a shared slot, and general upward/downward packets and DAO messages are transmitted from a dedicated cell to a neighbor node. Although the increase of packet overheads in the proposed scheme can induce collisions and an increase in energy usage, the control message and the retransmission message overhead, which are generated in the proposed scheme, are not that high as compared to standard RPL.

Experiment using a simulator

The OpenWSN simulator is used to compare the performance of IM-RPL against standard RPL to evaluate performance in less constrained topologies for longer periods of time.

The propagation model of the OpenWSN simulator is based on the Pister-hack model.³⁹ The RSSI is calculated by subtracting a uniform random value between 0 and 40 dB from a value derived from the Friis model, a free-space radio loss model. In this simulation, the TX power of all nodes is set to −30 dBm. Figure 19 depicts a measure of the ratio of packet reception according to the distance between two nodes using the simulator. When the distance between the two nodes is within 2 m, more than 80% of the PRR is shown, but it is lowered below 50% at more than 5 m and 10% of PRR at 15 m.

Figure 19.

Packet reception ratio by distance when the TX power is −30 dBm in the OpenWSN simulator.

In each simulation, the root node sends User Datagram Protocol (UDP) packets to the MN for 5 h at a 30 s cycle to measure the PRR in the downward traffic. The maximum number of retransmissions of the MAC layer is 3, and the control message period is 60 s for the DAO message, 30 s for the DIO message, and 10 s for the DIS message of the MNs. The topology comprises topology_A and topology_B, as shown in Figure 20. Topology_A includes one root node, four static nodes, and one MN. The static nodes and root node are linearly located at 4 m intervals. The MN moves horizontally at 4 m intervals from the static nodes at speeds of 0.12, 0.24, and 0.36 m/s. Topology_B includes 1 root node, 12 static nodes, and 4 MNs. The MNs move between the static nodes at a speed of 0.12 m/s.

Figure 20.

(a) Topology_A and (b) topology_B in experiment using a simulator.

Performance evaluation

The two simulated topologies, topology_A and topology_B, are represented in Figure 20. PRR according to the speed of the MN was measured in topology_A, and the same was measured according to the number of MNs in topology_B.

Figure 21 shows the PRR at topology_A with MN speeds of 0.12, 0.24, and 0.36 m/s. The standard RPL values are 29%, 24%, and 20%, and the IM-RPL values are 78%, 79%, and 77%, respectively. In the standard RPL, the speed of the MN increased from 0.12 to 0.36 m/s, resulting in a 9% decrease in PRR and a large change depending on the speed of the MN. The faster the MN, the more often the parent switch occurs. The MN needs to receive more than a certain number of packets from the neighbor nodes to perform the parent switch, and it takes time to update the downward routing information of the root node even after the parent switch occurs. However, in the case of IM-RPL, even if the speed of the MN is increased to 0.36 m/s, the PRR does not vary significantly. IM-RPL is very reliable for downward traffic by updating the routing information quickly and forwarding retransmission packets through neighboring nodes even if the MN exits the transmission range. However, if the MN becomes faster, the PRR in IM-RPL will also decrease.

Figure 21.

Packet reception ratio in topology_A.

Figure 22 presents the PRR when the number of MNs moving at a constant speed of 0.12 m/s in topology_B is 1, 2, and 4. The PRRs in standard RPL are 22%, 24%, and 20%, and those in IM-RPL are 96%, 96%, and 94%, respectively. In both standard RPL and IM-RPL, increasing the number of MNs has no significant effect on PRR. Rather, incorrect routing information will have a significant effect on the PRR of downward traffic. In IM-RPL, even if the root node sends a downward packet in the wrong path, it can be delivered to the correct destination through the retransmission method, and it records very high PRR values because it updates the information quickly with the quick parent-change method.

Figure 22.

Packet reception ratio in topology_B.

In Figure 22, the PRR for IM-RPL is higher than that in Figure 21. This is because a node in topology_B has more neighboring nodes than one in topology_A and nodes in topology_B can forward more broadcasted packets. This creates a higher probability of receiving a packet.

Unlike in the simulation environment, the difference in PRR by the number of MNs will be larger in the real world owing to greater radio interference.

Overhead evaluation

The overheads are compared in the same simulation. The RPL control messages DIO, DAO, and DIS transmitted are defined as overheads. In the case of IM-RPL, the broadcast messages for retransmission and the packets that forward the broadcasted message to the MN are also defined as overheads.

Figure 23 shows the total overhead incurred in topology_A. In each scheme, the overhead difference according to the speed of the MN is not large, but the overhead in IM-RPL is approximately 19%–22% higher than that in the standard RPL. This is because, in IM-RPL, the MN sends more DAOs as soon as they switch their parent, and the retransmission scheme incurs the additional overhead. However, this overhead is not as much as in the case of PRRs, which increase by approximately 48%–56%.

Figure 23.

Number of total overheads in topology_A.

Figure 24 presents the total overhead incurred in topology_B. In both standard RPL and IM-RPL, the greater the number of MNs, the greater is the overhead. This is because the number of DIOs increases owing to the DIS transmissions of the MNs. The overhead in IM-RPL is approximately 9%–30% higher than that in standard RPL because of the added overhead due to the IM-RPL retransmission scheme.

Figure 24.

Number of total overheads in topology_B.

The results in the simulations are similar to those in the experiments using hardware devices. IM-RPL recorded much higher PRRs with little overhead compared to the existing RPL. Owing to the poor performance of the OpenWSN simulator, the same could not be simulated in a large topology with very many nodes.

Conclusion and future work

This article discusses the problems caused by MNs in RPL and proposes a retransmission technique that can increase the performance of downward traffic to the MNs by using control packets, neighbor nodes, and a fast parent-change support technique in a dynamic environment. Experiments and simulations were performed in OpenWSN’s TSCH. The proposed scheme has high reliability and low transmission delay as compared to the standard RPL. The performance evaluation using hardware devices reveals that the MN should not exceed the limit speed for the proposed scheme. In addition, the control message and the broadcast overheads for retransmission generated by the proposed scheme were measured. It indicated a PRR improvement and a low transmission delay time based on the appropriate overhead message. This can address various mobility problems, applications, and user QoS aspects required in industrial environments.

Although the conducted experiments were based on the non-storing mode, the proposed method can be applied to the storing mode as well. Further experiments in this area will involve implementing the storing mode in OpenWSN’s TSCH environment and experimenting based on each mode’s characteristics. This will help identify and test the real-time environment applications that must be performed in a certain time period in an industrial environment, as well as experiments on a plant or factory.

Footnotes

Handling Editor: Carlos Calafate

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research,authorship,and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research,authorship,and/or publication of this article: This work was supported by the Korea Institute of Energy Technology Evaluation and Planning (KETEP) and the Ministry of Trade,Industry & Energy (MOTIE) of the Republic of Korea (No. 20182020109660). This research was supported by the MSIT (Ministry of Science and ICT),Korea,under the ICT Consilience Creative program (IITP-2019-2016-0-00318) supervised by the IITP (Institute for Information & communications Technology Planning & Evaluation).

ORCID iDs

Soon-Woong Min

Hee-Jun Lee

References

Lasi

Fettke

Kemper

H-G

, et al. Industry 4.0. Bus Inform Syst Eng 2014; 6(4): 239–242.

Wan

, et al. A review of industrial wireless networks in the context of industry 4.0. Wirel Netw 2017; 23(1): 23–41.

Zuehlke

Smart factory—towards a factory-of-things. Annu Rev Contol 2010; 34: 129–138.

Petersen

Carlsen

Wireless HART versus ISA100.11a: the format war hits the factory floor. IEE Ind Electron M 2011; 5(4): 23–34.

De Guglielmo

. IEEE 802.15.4e: a survey. Comput Commun 2016; 88: 1–24.

Sisinni

Industrial internet of things: challenges, opportunities, and directions. IEEE T Ind Inform 2018; 14(11): 4724–4734.

Jain

Networking protocols for internet of things. St. Louis, MO: Washington University in St. Louis, 2013.

Winter

Thubert

. RPL: IPv6 routing protocol for low-power and lossy networks. RFC6550, 2012, https://tools.ietf.org/html/rfc6550

Omprakash

Fonseca

RLC

Jamieson

, et al. CTP: an efficient robust reliable collection tree protocol for wireless sensor networks. ACM T Sensor Network 2013; 10(1): 16.

10.

Carels

De Poorter

. RPL mobility support for point-to-point traffic flows towards mobile nodes. Int J Distrib Sens N 2015; 10(1): 470349.

11.

Breivold

Sandström

. Internet of things for industrial automation—challenges and technical solutions. In: 2015 IEEE international conference on data science and data intensive systems, Sydney, NSW, Australia, 11–13 December 2015. New York: IEEE.

12.

Kim

H-S

Culler

, et al. Challenging the IPv6 routing protocol for low-power and lossy networks (RPL): a survey. IEEE Commun Surv Tutor 2017; 19(4): 2502–2525.

13.

Watteyne

Palattella

. Using IEEE 802.15.4e time-slotted channel hopping (TSCH) in the Internet of Things (IoT): problem statement. RFC7554, 2015, https://tools.ietf.org/html/rfc7554

14.

Mohammad

Guo

. Oppcast: exploiting spatial and channel diversity for robust data collection in urban environments. In: IPSN ‘16 proceedings of the 15th international conference on information processing in sensor networks, Vienna, 11–14 April 2016. New York: IEEE.

15.

Manogaran

Varatharajan

Lopez

, et al. A new architecture of Internet of Things big data ecosystem for secured smart healthcare monitoring alerting system. Future Gener Comp Sy 2018; 82: 375–387.

16.

Nassar

Berthomé

Dubrulle

, et al. Multiple instances QoS routing in RPL: application to smart grids. Sensors 2018; 18(8): E2472.

17.

Gaddour

Koubâa

. OF-FL: QoS-aware fuzzy logic objective function for the RPL routing protocol. In: 2014 12th international symposium on modeling and optimization in mobile, ad hoc, and wireless networks (WiOpt), Hammamet, Tunisia, 12–16 May 2014. New York: IEEE.

18.

Harshavardhana

Vineeth

Anand

SVR

, et al. Power control and cross-layer design of RPL objective function for low power and lossy networks. In: 2018 10th international conference on communication systems & networks (COMSNETS), Bengaluru, India, 3–7 January 2018. New York: IEEE.

19.

Mardini

Aljawarneh

Al-Abdi

, et al. Performance evaluation of RPL objective functions for different sending intervals. In: 2018 6th international symposium on digital forensic and security (ISDFS), Antalya, Turkey, 22–25 March 2018. New York: IEEE.

20.

Lamaazi

Benamar

. RPL enhancement using a new objective function based on combined metrics. In: 2017 13th international wireless communications and mobile computing conference (IWCMC), Valencia, 26–30 June 2017. New York: IEEE.

21.

Sousa

Sobral

JVV

Rodrigues

JJPC

, et al. ERAOF: a new RPL protocol objective function for Internet of Things applications. In: 2017 2nd international multidisciplinary conference on computer and energy science (SpliTech), Split, 12–14 July 2017. New York: IEEE.

22.

Lamaazi

Benamar

Jara

. Study of the impact of designed objective function on the RPL-based routing protocol. Adv Ubiquit Network 2016; 397: 67–80.

23.

Lemercier

Montavont

. A new objective function for hybrid network in the smart grid. In: 2018 IEEE 19th international symposium on “a world of wireless, mobile and multimedia networks” (WoWMoM), Chania, 12–15 June 2018. New York: IEEE.

24.

Chen

Chanet

J-P

Hou

K-M

, et al. A scalable context-aware objective function (SCAOF) of routing protocol for agricultural low-power and lossy networks (RPAL). Sensors 2015; 15: 19507–19540.

25.

Duquennoy

Eriksson

Voigt

. Five-nines reliable downward routing in RPL. Computer Science Networking and Internet Architecture, 2017, https://arxiv.org/abs/1710.02324.

26.

Hyung-Sin

Kim

. DT-RPL: Diverse bidirectional traffic delivery through RPL, routing protocol in low power lossy. Comput Netw 2017; 126(24): 150–161.

27.

Gardašević

Veletić

. The IoT architectural framework design issues application domains. Wireless Pers Commun 2017; 92(1): 127–148.

28.

Korbi

Ben Brahim

Adjih

. Mobility enhanced RPL for wireless sensor networks. In: Network of the future (NOF) 2012 third international conference, Gammarth, Tunisia, 21–23 November 2012, pp.1–8. New York: IEEE.

29.

Chang

. MoMoRo: providing mobility support for low-power wireless applications. IEEE Sys J 2015; 9(2): 585–594.

30.

Gaddour

Koubaa

Rangarajan

. CoRPL: RPL routing for mobile low power wireless sensor networks using corona mechanism. In: 2014 9th IEEE international symposium industrial embedded systems (SIES), Pisa, 18–20 June 2014, pp.200–209. New York: IEEE.

31.

Fotouhi

Moreira

. MRPL: boosting mobility in the internet of things. Ad Hoc Network 2015; 26: 17–35.

32.

Ancillotti

Vallati

Bruno

, et al. A reinforcement learning-based link quality estimation strategy for RPL and its impact on topology management. Comput Commun 2017; 112: 1–13.

33.

Sanshi

Jaidhar

. Enhanced mobility aware routing protocol for Low Power and Lossy Networks. Wireless Netw 2017; 25: 1641–1655.

34.

Kharrufa

Al-Kashoash

Kemp

. A game theoretic optimization of RPL for mobile internet of things applications. IEEE Sensor J 2018; 18: 2520–2530.

35.

Bouaziz

Rachedi

Belghith

, et al. EMA-RPL: energy and mobility aware routing for the Internet of Mobile Things. Future Gener Comp Sy 2019; 97: 247–258.

36.

Levis

Clausen

. The trickle algorithm. RFC6206, 2011, https://tools.ietf.org/html/rfc6206

37.

https://openwsn.atlassian.net/

38.

http://www.openmote.com/

39.

John

Pister

. Energy-aware routing in wireless sensor networks with adaptive energy-slope control. EE290Q-2 Spring, 2009, https://s3.amazonaws.com/academia.edu.documents/3471034/EE290_PhucMervin.pdf?response-content-disposition=inline%3B%20filename%3DEnergy-Aware_Routing_in_Wireless_Sensor.pdf&X-Amz-Algorithm=AWS4-HMAC-SHA256&X-Amz-Credential=AKIAIWOWYYGZ2Y53UL3A%2F20200128%2Fus-east-1%2Fs3%2Faws4_request&X-Amz-Date=20200128T094225Z&X-Amz-Expires=3600&X-Amz-SignedHeaders=host&X-Amz-Signature=7fa775783a41a1f3d4004a8165f33ccf28635023b2fe63c398cbcb8d09b72fad

Downward traffic retransmission mechanism for improving reliability in RPL environment supporting mobility

Abstract

Keywords

Introduction

Background

802.15.4e: TSCH

RPL

Constructing DODAG for routing

Storing/non-storing mode

Related work

Upward/downward traffic

Mobility

Downward traffic problems and mobility support technique

Upward/downward problem

Mobility problem

Downward traffic support scheme under mobility

Quick parent-change support scheme

Analysis of packet overhead

Experiment

Experiment using hardware devices

Resynchronization and routing update delay

Performance evaluation

Overhead evaluation

Experiment using a simulator

Performance evaluation

Overhead evaluation

Conclusion and future work

Footnotes

Declaration of conflicting interests

Funding

ORCID iDs

References